Benchmark Almanac

Decision Index

karpo-decision-31B 73.9 tops its leaderboard Gemma 4 31B IT 67.6 leads the models we track
Not in model cards

Decision models read a state and a typed question and return a decision. The index weights private same-skill sets and new-domain sets, equated on 21 base models.

Tracks reproductions of TypeSafe's Jev decision model. Version 0.4.1 board lists 248 entries. A separate vision board exists.

Released
September 2026
Kind
Composite indices
Level
Everyday
Scored
Checked automatically
Answers
Prediction
Tasks
Public, plus a private test set
Strongest model tested
Qwen 3.8 27B #58 on the capability index
Last model added
August 2026
Scores refreshed
November 12th, 2026
297
Entries on its own leaderboard, 3 models we track

karpo-decision-31B tops the leaderboard with 73.9

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Gemma 4 31B IT, at 67.6.

The newest model with a score here came out in August 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

297 entries. Scroll sideways for the rest.

  1. 73.9 karpo-decision-31B
  2. 73.4 GLiDE 27B
  3. 72.6 REAX 1.3
  4. 72.3 Logits Sigma-1 Duo
  5. 72.1 Copernicus-31B
  6. 71.9 KnowLine 31B Gen1.5
  7. 71.8 Copernicus-57B
  8. 71.6 FlexDM-LoRA 2.1 · Gemma-4-31B
  9. 71.1 Sieve 31B v3
  10. 70.9 OneJev1.6g-31B
  11. 70.5 Turion Jev 31B
  12. 70.3 eshbar-lulilla (Gemma 4 31B)
  13. 70.3 Jeffan-27B-1211
  14. 70.2 vLLM-SR d3-ultra (Decision 3.0, 27B)
  15. 70.0 Bespoke Nimble v4.1 Max (27B)
  16. 69.9 H2O Lightning 31B v1.3.0
  17. 69.8 Jeffan-31b-1710
  18. 69.5 Logits Sigma-1
  19. 69.1 liquamen (Gemma 4 31B)
  20. 69.0 Cerno 27B v1.1
  21. 68.5 WeMM-JEV-27B
  22. 68.4 baselabs mammoth-1 (27B)
  23. 68.4 Piccolo 27B
  24. 68.2 Ainvest-Decision-27B v1.0
  25. 68.1 oenogarum (Gemma 4 31B)
  26. 68.0 Perplexity Decider v1.2 (27B)
  27. 68.0 AutoTrust-JEV-27B v1.1
  28. 67.9 StartLux-Decision v1.1 27B
  29. 67.7 Perplexity Decider v1.3 (27B)
  30. 67.6 MISA AI 2.0 Decision Plus · oak 1.1 (31B)
  31. 67.6 AJev lm6q (Gemma 4 31B)
  32. 67.6 Gemma 4 31B IT
  33. 67.5 Chon-1.1-31B
  34. 67.5 decy-preview (31B)
  35. 67.3 Lychee-27B
  36. 67.3 vLLM-SR d3-turbo v1.1 (27B)
  37. 67.2 vLLM-SR d3 (27B)
  38. 67.0 allec-oavg (Gemma 4 31B)
  39. 66.9 VisionJEV 1.0.0 (27B)
  40. 66.7 Aleria Judge 1.1.1 (Gemma 4 31B)
  41. 66.6 Nace.AI Drex Max v1 (27B)
  42. 66.5 AnyJev Tacitus 31B
  43. 66.3 vera-large-2o
  44. 66.0 Decider 31B (NVFP4)
  45. 65.8 KnowLine 35B-A3B Gen1.5
  46. 65.7 OneShot 27B
  47. 65.2 Valen-Decision 1.5 27B
  48. 64.8 Bobcat 1.3 (31B)
  49. 64.8 Amos 1 Large (27B)
  50. 64.8 Matilda Decision v1
  51. 64.6 KnowLine 26B-A4B Gen2.5
  52. 64.4 frontier-infra Jebadiah 27B v2.1 (t1e)
  53. 64.2 Darwin-27B-RSI-Decision
  54. 63.0 deck31b (Gemma 4 31B)
  55. 62.7 decisio 0.6.0 (Gemma 4 31B)
  56. 62.5 GLiDE 9B
  57. 61.5 Torchcast Decision 27B
  58. 61.5 Jeffan-9B-1211
  59. 61.5 OneJev1.6-27B
  60. 61.1 Qwen Decision 27B noncausal LoRA
  61. 60.8 Quyet-1.0-Large
  62. 60.3 garum (Gemma 4 31B)
  63. 60.3 baselabs elephant-1 (9B)
  64. 60.1 Standard One 35B SH (v1)
  65. 60.1 NTele-Hybrid-27B-v1
  66. 60.0 Lumma-fev-27B
  67. 59.9 Sieve 27B v2
  68. 59.8 mogan-decision-31B-it
  69. 59.6 linne-fern 1.0 (27B)
  70. 59.6 Kev 27B
  71. 59.4 Xor 26B-A4B v1.2
  72. 59.2 Jev 1.13
  73. 59.2 Adjudex-9B v5
  74. 59.1 span-02 (Gemma 4 26B-A4B)
  75. 59.0 Cerno 9B v1.2
  76. 59.0 garas (Gemma 4 31B)
  77. 58.8 Shisa DE-2.1 (repeat)
  78. 58.8 Eikos-27B
  79. 58.8 reflex Qwen3.8-27B-FP8 (pmscr)
  80. 58.6 RSI-Jev v6.1 VL 27B (effort auto)
  81. 58.4 KnowLine 9B Gen1
  82. 58.4 H2O Lightning Bolt 74B v1.1
  83. 58.1 StartLux-Decision v1.1 9B
  84. 58.0 vLLM-SR Decision 2.0 Vega 27B
  85. 57.6 JEV-35B
  86. 57.4 JEV-27B
  87. 57.3 Piccolo 9B
  88. 57.0 Qwen 3.8 27B
  89. 56.6 Tern-35B-A3B
  90. 56.4 Surogate Rune 26B-A4B v3
  91. 56.2 SynACK Decide 26B-A4B v7
  92. 56.0 KnowLine 4B Gen5.5
  93. 55.9 Blink v0.3 26B-A4B (NVFP4)
  94. 55.9 Valen-Decision 1.5 9B
  95. 55.6 JADE
  96. 55.4 Dexy-Decision-9B-V2
  97. 55.4 Wald-26B-A4B
  98. 55.3 AJev lora1 (Gemma 4 26B-A4B)
  99. 55.1 Perplexity Decider Flash v1.3 (4B)
  100. 54.9 Cloudflare clef
  101. 54.8 StartLux-Decision v1.1 4B
  102. 54.7 Qwen3.6 27B
  103. 54.3 Perplexity Decider Flash v1.2 (4B)
  104. 54.3 vLLM-SR d3-flash (9B)
  105. 54.1 JPT-35B-A3B
  106. 54.0 chinese-decision-pointer-4b
  107. 53.9 Cerno 4B v1.2
  108. 53.6 Xor 1.2
  109. 53.5 GEV-26B-Decide
  110. 53.0 OneJev1.6-9B
  111. 53.0 vera-medium-omni
  112. 52.8 Senova-4B v4
  113. 52.8 Nace.AI Drex v2.0 (35B-A3B)
  114. 52.5 muria (Gemma 4 12B)
  115. 52.2 Jeffan-4B-2010
  116. 52.0 Verdin-4B v1.1
  117. 51.9 LatentMesh Blitz 4B v1
  118. 51.7 Saif 4B v2
  119. 51.1 vev-4b-v2
  120. 51.1 Bespoke Nimble 9B v3
  121. 50.9 frontier-infra Jebadiah 9B v2.1 (t1e)
  122. 50.8 decisio 0.4.0 (Qwen3.6-35B-A3B)
  123. 50.0 Piccolo 4B
  124. 50.0 vLLM-SR d3-mini (4B)
  125. 49.9 RSI-Jev v7.0 VL 4B
  126. 49.9 Valen-Decision 1.5 4B
  127. 49.8 Jegue 0.2 (9B)
  128. 49.7 RSI-Jev v7.0 VL 3B
  129. 49.5 JoshuaSP diffusiongemma (open-jev)
  130. 48.9 AJev lora7 (Gemma 4 12B)
  131. 48.9 OneJev1.6-4B
  132. 48.8 decisio 0.4.0 (Gemma 4 12B)
  133. 48.5 BaseDecision-Dyno (4B)
  134. 48.0 H2O Lightning E4B v1.0.1
  135. 47.7 Mica v0.5 4B
  136. 47.5 PurpleMIST-Flash-1.0
  137. 47.0 Sieve 9B v2
  138. 46.6 Decider 35B-A3B
  139. 46.6 Winnow-12B
  140. 46.3 Seb-9B v2
  141. 46.2 PurpleMIST-Mini-2.0
  142. 46.0 vLLM-SR Decision 2.0 Lux 9B
  143. 45.8 Nace.AI Drex v1.5
  144. 45.8 Hopper 12B (trained)
  145. 45.6 Trio-Spark DI (9B)
  146. 45.6 Jevora 9B v1.6
  147. 45.5 CoCo-Decision-4B 2.4
  148. 45.0 Cloudflare clef-omni (30B-A3B)
  149. 44.9 JevK5 9B v0.3.3
  150. 44.4 Cloudflare clef-flash
  151. 44.4 Jevfire
  152. 44.1 jiwo 4B v2
  153. 44.1 JPT-9B
  154. 43.9 ezjev 4B s2
  155. 43.8 Jev-Omni
  156. 43.7 GLiDE 2B
  157. 43.2 Solomon v1.1
  158. 42.9 Standard One 8B SH (v4)
  159. 42.7 Amber Hybrid 4B
  160. 42.1 JEV-9B
  161. 42.1 Bespoke Nimble 9B v2
  162. 41.7 Sieve 9B
  163. 41.0 Sieve 4B
  164. 40.9 Kev 9B v2
  165. 40.9 Nari Decision 4B v3.2
  166. 40.8 Aplomb 1
  167. 40.8 vLLM-SR Decision 2.0 Nox 4B
  168. 40.2 Hopper (G) 1.2
  169. 40.1 StartLux-Decision v1.1 2B
  170. 39.8 FlyMy.AI Decision 4B v1.6
  171. 39.8 Standard One 3B SH (v4)
  172. 39.6 JPT-4B
  173. 39.5 Amber 4B
  174. 38.7 Intern-Decision-4B
  175. 38.1 djev
  176. 38.1 KnowLine 2B Gen1
  177. 38.0 Valen-Decision 1.5 2B
  178. 37.6 razorback16 openjev diffusiongemma (NVFP4, vLLM)
  179. 37.3 NeoHorse-Jev-4B
  180. 36.6 JevK5 v0.3
  181. 36.4 GLiDE 0.8B
  182. 36.3 Decider 4B
  183. 35.9 baselabs joey-1 (2B)
  184. 35.5 lev
  185. 35.4 Lev1 3.7B
  186. 35.1 Jev-Style 2B v4.2
  187. 34.9 Kev 4B r10
  188. 34.8 PKU-KCL-Meteor-2B
  189. 34.5 Sieve 2B v3
  190. 34.4 LatentMesh Blitz 2B v1
  191. 34.4 Saracura PT-BR 4B
  192. 34.4 LiquidAI d1-2B
  193. 34.4 Cam1
  194. 34.3 vLLM-SR d3-nano (2B)
  195. 34.3 mmastrac diffusiongemma (vLLM PR 57250)
  196. 34.0 Kas 4B
  197. 33.8 GLiDE 0.4B
  198. 33.7 Tev1-4B-experimental
  199. 33.6 LiquidAI d1-Omni-2B
  200. 32.8 Adjudex-0.8B v5
  201. 32.7 Winnow-E4B
  202. 32.5 Aurelis-2B
  203. 32.5 GLiClass v4.1 edge (0.4B)
  204. 32.2 Jevstral 8B
  205. 32.2 Metask-Jev-4B
  206. 32.0 Ryotide Qwen v0.3.1
  207. 31.3 Intelif Qwen3-4B
  208. 31.1 LiquidAI d1-3B
  209. 31.1 open-jev (pngwn)
  210. 30.9 PKU-KCL-Meteor-0.8B
  211. 30.6 vera-core-2
  212. 30.4 Jobe
  213. 29.6 OneJev1.6-0.8B
  214. 29.4 crystal-0.8b
  215. 29.4 MoDe-L
  216. 29.3 vLLM-SR vela3-fast (Vela 3.0, 0.4B)
  217. 29.3 Crivo-1B
  218. 29.2 KnowLine 0.8B Gen1
  219. 29.2 Gevva E4B
  220. 28.9 StartLux-Decision v1.1 0.8B
  221. 28.8 openvons
  222. 27.8 Amber 0.4B v2
  223. 27.6 Valen-Decision 1.5 0.8B
  224. 26.7 SemIf
  225. 26.5 Saina Helm 2 Omni (0.8B)
  226. 26.5 MoDe-M
  227. 26.4 Saif 0.8B Lightning v0.3
  228. 26.2 vLLM-SR d3-lite (0.8B)
  229. 25.8 jiwo 0.8B v2
  230. 25.3 Saina Helm 2.0 v2
  231. 24.9 this-that 1.2
  232. 24.7 Amber 150M
  233. 24.5 MoDe-S
  234. 24.3 Gevva E2B
  235. 23.7 Crivo-400M
  236. 23.6 Decider 2B
  237. 23.6 vLLM-SR Decision 2.0 Sol 2B
  238. 23.6 jdi-ptr14-0.8b
  239. 22.8 mini-jev
  240. 22.1 vLLM-SR vela3-fast-vision (Vela 3.0, 0.25B)
  241. 21.9 Sieve 2B
  242. 21.5 wyat-1-800m
  243. 20.9 Intern-Decision-2B
  244. 20.8 Sieve 0.8B
  245. 20.7 Strands Decider 2B
  246. 20.6 EXAONE-4.0-1.2B-JEV v0.3
  247. 19.9 tasksource Decider nano v5 (150M)
  248. 19.5 Crivo-68M
  249. 19.2 Amber 68M
  250. 18.5 JEV Ettin 150M
  251. 18.4 vLLM-SR Decision 2.0 Eos 0.8B
  252. 18.3 Crivo-150M v2
  253. 18.0 Bekko System One v0 400M
  254. 17.8 Mola-150M
  255. 17.7 vLLM-SR d3-edge (0.6B)
  256. 17.6 Mola-68M
  257. 17.3 JEV FLAN-T5 Large public v2 (342M)
  258. 17.2 DecisionTune 1.0
  259. 16.5 Bosun v3.1 1.7B
  260. 16.4 vLLM-SR Decision 2.0 Kai 0.6B
  261. 16.4 Dinah-1
  262. 16.3 Kev 0.8B r15
  263. 16.0 Amber 32M
  264. 15.6 PICK-150M v5
  265. 15.5 JPT-0.8B
  266. 15.2 JEV Ettin 400M
  267. 14.5 Sifr 0.8B v3.1
  268. 14.2 Crivo-32M v2
  269. 14.0 Prumo
  270. 13.9 Kodiak v0.2 1B (accuracy)
  271. 12.7 Intern-Decision-0.8B
  272. 12.5 Jev-Style 0.8B v3
  273. 12.5 Tev1-0.8B-experimental
  274. 12.4 Bekko System One v0 68M
  275. 12.0 Amber 17M
  276. 11.8 LiquidAI d1-omni-600M
  277. 11.7 Crivo-17M
  278. 11.7 Von 1.3
  279. 11.0 Bosun v3.1 0.6B
  280. 11.0 vera-spark
  281. 10.8 Lavoir
  282. 10.1 GLiNER2.5-Decide
  283. 9.2 Laya
  284. 8.8 JevK5 Lite
  285. 8.2 LFM2.5-2.6B-RLCD
  286. 7.8 Qwen-2.5-1B-RLCD
  287. 7.7 Bekko System One v0 17M
  288. 7.4 EmbeddingGemma 2 System-One
  289. 6.5 CLM-v0.1-8B
  290. 6.2 jeff
  291. 5.9 LFM2.5-350M-RLCD
  292. 5.2 system-one-gemma
  293. 5.0 DM-JEPA 1.1
  294. 4.8 Julia 1
  295. 3.5 Nirnaya 0.4B
  296. 3.4 Zev
  297. 2.0 Verdict
  1. 1 karpo-decision-31B 73.9
  2. 2 GLiDE 27B 73.4
  3. 3 REAX 1.3 72.6
  4. 4 Logits Sigma-1 Duo 72.3
  5. 5 Copernicus-31B 72.1
  6. 6 KnowLine 31B Gen1.5 71.9
  7. 7 Copernicus-57B 71.8
  8. 8 FlexDM-LoRA 2.1 · Gemma-4-31B 71.6
  9. 9 Sieve 31B v3 71.1
  10. 10 OneJev1.6g-31B 70.9
  11. 11 Turion Jev 31B 70.5
  12. 12 eshbar-lulilla (Gemma 4 31B) 70.3
  13. 13 Jeffan-27B-1211 70.3
  14. 14 vLLM-SR d3-ultra (Decision 3.0, 27B) 70.2
  15. 15 Bespoke Nimble v4.1 Max (27B) 70.0
  16. 16 H2O Lightning 31B v1.3.0 69.9
  17. 17 Jeffan-31b-1710 69.8
  18. 18 Logits Sigma-1 69.5
  19. 19 liquamen (Gemma 4 31B) 69.1
  20. 20 Cerno 27B v1.1 69.0
  21. 21 WeMM-JEV-27B 68.5
  22. 22 baselabs mammoth-1 (27B) 68.4
  23. 23 Piccolo 27B 68.4
  24. 24 Ainvest-Decision-27B v1.0 68.2
  25. 25 oenogarum (Gemma 4 31B) 68.1
  26. 26 Perplexity Decider v1.2 (27B) 68.0
  27. 27 AutoTrust-JEV-27B v1.1 68.0
  28. 28 StartLux-Decision v1.1 27B 67.9
  29. 29 Perplexity Decider v1.3 (27B) 67.7
  30. 30 MISA AI 2.0 Decision Plus · oak 1.1 (31B) 67.6
  31. 31 AJev lm6q (Gemma 4 31B) 67.6
  32. 32 Gemma 4 31B IT 67.6
  33. 33 Chon-1.1-31B 67.5
  34. 34 decy-preview (31B) 67.5
  35. 35 Lychee-27B 67.3
  36. 36 vLLM-SR d3-turbo v1.1 (27B) 67.3
  37. 37 vLLM-SR d3 (27B) 67.2
  38. 38 allec-oavg (Gemma 4 31B) 67.0
  39. 39 VisionJEV 1.0.0 (27B) 66.9
  40. 40 Aleria Judge 1.1.1 (Gemma 4 31B) 66.7
  41. 41 Nace.AI Drex Max v1 (27B) 66.6
  42. 42 AnyJev Tacitus 31B 66.5
  43. 43 vera-large-2o 66.3
  44. 44 Decider 31B (NVFP4) 66.0
  45. 45 KnowLine 35B-A3B Gen1.5 65.8
  46. 46 OneShot 27B 65.7
  47. 47 Valen-Decision 1.5 27B 65.2
  48. 48 Bobcat 1.3 (31B) 64.8
  49. 49 Amos 1 Large (27B) 64.8
  50. 50 Matilda Decision v1 64.8
  51. 51 KnowLine 26B-A4B Gen2.5 64.6
  52. 52 frontier-infra Jebadiah 27B v2.1 (t1e) 64.4
  53. 53 Darwin-27B-RSI-Decision 64.2
  54. 54 deck31b (Gemma 4 31B) 63.0
  55. 55 decisio 0.6.0 (Gemma 4 31B) 62.7
  56. 56 GLiDE 9B 62.5
  57. 57 Torchcast Decision 27B 61.5
  58. 58 Jeffan-9B-1211 61.5
  59. 59 OneJev1.6-27B 61.5
  60. 60 Qwen Decision 27B noncausal LoRA 61.1
  61. 61 Quyet-1.0-Large 60.8
  62. 62 garum (Gemma 4 31B) 60.3
  63. 63 baselabs elephant-1 (9B) 60.3
  64. 64 Standard One 35B SH (v1) 60.1
  65. 65 NTele-Hybrid-27B-v1 60.1
  66. 66 Lumma-fev-27B 60.0
  67. 67 Sieve 27B v2 59.9
  68. 68 mogan-decision-31B-it 59.8
  69. 69 linne-fern 1.0 (27B) 59.6
  70. 70 Kev 27B 59.6
  71. 71 Xor 26B-A4B v1.2 59.4
  72. 72 Jev 1.13 59.2
  73. 73 Adjudex-9B v5 59.2
  74. 74 span-02 (Gemma 4 26B-A4B) 59.1
  75. 75 Cerno 9B v1.2 59.0
  76. 76 garas (Gemma 4 31B) 59.0
  77. 77 Shisa DE-2.1 (repeat) 58.8
  78. 78 Eikos-27B 58.8
  79. 79 reflex Qwen3.8-27B-FP8 (pmscr) 58.8
  80. 80 RSI-Jev v6.1 VL 27B (effort auto) 58.6
  81. 81 KnowLine 9B Gen1 58.4
  82. 82 H2O Lightning Bolt 74B v1.1 58.4
  83. 83 StartLux-Decision v1.1 9B 58.1
  84. 84 vLLM-SR Decision 2.0 Vega 27B 58.0
  85. 85 JEV-35B 57.6
  86. 86 JEV-27B 57.4
  87. 87 Piccolo 9B 57.3
  88. 88 Qwen 3.8 27B 57.0
  89. 89 Tern-35B-A3B 56.6
  90. 90 Surogate Rune 26B-A4B v3 56.4
  91. 91 SynACK Decide 26B-A4B v7 56.2
  92. 92 KnowLine 4B Gen5.5 56.0
  93. 93 Blink v0.3 26B-A4B (NVFP4) 55.9
  94. 94 Valen-Decision 1.5 9B 55.9
  95. 95 JADE 55.6
  96. 96 Dexy-Decision-9B-V2 55.4
  97. 97 Wald-26B-A4B 55.4
  98. 98 AJev lora1 (Gemma 4 26B-A4B) 55.3
  99. 99 Perplexity Decider Flash v1.3 (4B) 55.1
  100. 100 Cloudflare clef 54.9
  101. 101 StartLux-Decision v1.1 4B 54.8
  102. 102 Qwen3.6 27B 54.7
  103. 103 Perplexity Decider Flash v1.2 (4B) 54.3
  104. 104 vLLM-SR d3-flash (9B) 54.3
  105. 105 JPT-35B-A3B 54.1
  106. 106 chinese-decision-pointer-4b 54.0
  107. 107 Cerno 4B v1.2 53.9
  108. 108 Xor 1.2 53.6
  109. 109 GEV-26B-Decide 53.5
  110. 110 OneJev1.6-9B 53.0
  111. 111 vera-medium-omni 53.0
  112. 112 Senova-4B v4 52.8
  113. 113 Nace.AI Drex v2.0 (35B-A3B) 52.8
  114. 114 muria (Gemma 4 12B) 52.5
  115. 115 Jeffan-4B-2010 52.2
  116. 116 Verdin-4B v1.1 52.0
  117. 117 LatentMesh Blitz 4B v1 51.9
  118. 118 Saif 4B v2 51.7
  119. 119 vev-4b-v2 51.1
  120. 120 Bespoke Nimble 9B v3 51.1
  121. 121 frontier-infra Jebadiah 9B v2.1 (t1e) 50.9
  122. 122 decisio 0.4.0 (Qwen3.6-35B-A3B) 50.8
  123. 123 Piccolo 4B 50.0
  124. 124 vLLM-SR d3-mini (4B) 50.0
  125. 125 RSI-Jev v7.0 VL 4B 49.9
  126. 126 Valen-Decision 1.5 4B 49.9
  127. 127 Jegue 0.2 (9B) 49.8
  128. 128 RSI-Jev v7.0 VL 3B 49.7
  129. 129 JoshuaSP diffusiongemma (open-jev) 49.5
  130. 130 AJev lora7 (Gemma 4 12B) 48.9
  131. 131 OneJev1.6-4B 48.9
  132. 132 decisio 0.4.0 (Gemma 4 12B) 48.8
  133. 133 BaseDecision-Dyno (4B) 48.5
  134. 134 H2O Lightning E4B v1.0.1 48.0
  135. 135 Mica v0.5 4B 47.7
  136. 136 PurpleMIST-Flash-1.0 47.5
  137. 137 Sieve 9B v2 47.0
  138. 138 Decider 35B-A3B 46.6
  139. 139 Winnow-12B 46.6
  140. 140 Seb-9B v2 46.3
  141. 141 PurpleMIST-Mini-2.0 46.2
  142. 142 vLLM-SR Decision 2.0 Lux 9B 46.0
  143. 143 Nace.AI Drex v1.5 45.8
  144. 144 Hopper 12B (trained) 45.8
  145. 145 Trio-Spark DI (9B) 45.6
  146. 146 Jevora 9B v1.6 45.6
  147. 147 CoCo-Decision-4B 2.4 45.5
  148. 148 Cloudflare clef-omni (30B-A3B) 45.0
  149. 149 JevK5 9B v0.3.3 44.9
  150. 150 Cloudflare clef-flash 44.4
  151. 151 Jevfire 44.4
  152. 152 jiwo 4B v2 44.1
  153. 153 JPT-9B 44.1
  154. 154 ezjev 4B s2 43.9
  155. 155 Jev-Omni 43.8
  156. 156 GLiDE 2B 43.7
  157. 157 Solomon v1.1 43.2
  158. 158 Standard One 8B SH (v4) 42.9
  159. 159 Amber Hybrid 4B 42.7
  160. 160 JEV-9B 42.1
  161. 161 Bespoke Nimble 9B v2 42.1
  162. 162 Sieve 9B 41.7
  163. 163 Sieve 4B 41.0
  164. 164 Kev 9B v2 40.9
  165. 165 Nari Decision 4B v3.2 40.9
  166. 166 Aplomb 1 40.8
  167. 167 vLLM-SR Decision 2.0 Nox 4B 40.8
  168. 168 Hopper (G) 1.2 40.2
  169. 169 StartLux-Decision v1.1 2B 40.1
  170. 170 FlyMy.AI Decision 4B v1.6 39.8
  171. 171 Standard One 3B SH (v4) 39.8
  172. 172 JPT-4B 39.6
  173. 173 Amber 4B 39.5
  174. 174 Intern-Decision-4B 38.7
  175. 175 djev 38.1
  176. 176 KnowLine 2B Gen1 38.1
  177. 177 Valen-Decision 1.5 2B 38.0
  178. 178 razorback16 openjev diffusiongemma (NVFP4, vLLM) 37.6
  179. 179 NeoHorse-Jev-4B 37.3
  180. 180 JevK5 v0.3 36.6
  181. 181 GLiDE 0.8B 36.4
  182. 182 Decider 4B 36.3
  183. 183 baselabs joey-1 (2B) 35.9
  184. 184 lev 35.5
  185. 185 Lev1 3.7B 35.4
  186. 186 Jev-Style 2B v4.2 35.1
  187. 187 Kev 4B r10 34.9
  188. 188 PKU-KCL-Meteor-2B 34.8
  189. 189 Sieve 2B v3 34.5
  190. 190 LatentMesh Blitz 2B v1 34.4
  191. 191 Saracura PT-BR 4B 34.4
  192. 192 LiquidAI d1-2B 34.4
  193. 193 Cam1 34.4
  194. 194 vLLM-SR d3-nano (2B) 34.3
  195. 195 mmastrac diffusiongemma (vLLM PR 57250) 34.3
  196. 196 Kas 4B 34.0
  197. 197 GLiDE 0.4B 33.8
  198. 198 Tev1-4B-experimental 33.7
  199. 199 LiquidAI d1-Omni-2B 33.6
  200. 200 Adjudex-0.8B v5 32.8
  201. 201 Winnow-E4B 32.7
  202. 202 Aurelis-2B 32.5
  203. 203 GLiClass v4.1 edge (0.4B) 32.5
  204. 204 Jevstral 8B 32.2
  205. 205 Metask-Jev-4B 32.2
  206. 206 Ryotide Qwen v0.3.1 32.0
  207. 207 Intelif Qwen3-4B 31.3
  208. 208 LiquidAI d1-3B 31.1
  209. 209 open-jev (pngwn) 31.1
  210. 210 PKU-KCL-Meteor-0.8B 30.9
  211. 211 vera-core-2 30.6
  212. 212 Jobe 30.4
  213. 213 OneJev1.6-0.8B 29.6
  214. 214 crystal-0.8b 29.4
  215. 215 MoDe-L 29.4
  216. 216 vLLM-SR vela3-fast (Vela 3.0, 0.4B) 29.3
  217. 217 Crivo-1B 29.3
  218. 218 KnowLine 0.8B Gen1 29.2
  219. 219 Gevva E4B 29.2
  220. 220 StartLux-Decision v1.1 0.8B 28.9
  221. 221 openvons 28.8
  222. 222 Amber 0.4B v2 27.8
  223. 223 Valen-Decision 1.5 0.8B 27.6
  224. 224 SemIf 26.7
  225. 225 Saina Helm 2 Omni (0.8B) 26.5
  226. 226 MoDe-M 26.5
  227. 227 Saif 0.8B Lightning v0.3 26.4
  228. 228 vLLM-SR d3-lite (0.8B) 26.2
  229. 229 jiwo 0.8B v2 25.8
  230. 230 Saina Helm 2.0 v2 25.3
  231. 231 this-that 1.2 24.9
  232. 232 Amber 150M 24.7
  233. 233 MoDe-S 24.5
  234. 234 Gevva E2B 24.3
  235. 235 Crivo-400M 23.7
  236. 236 Decider 2B 23.6
  237. 237 vLLM-SR Decision 2.0 Sol 2B 23.6
  238. 238 jdi-ptr14-0.8b 23.6
  239. 239 mini-jev 22.8
  240. 240 vLLM-SR vela3-fast-vision (Vela 3.0, 0.25B) 22.1
  241. 241 Sieve 2B 21.9
  242. 242 wyat-1-800m 21.5
  243. 243 Intern-Decision-2B 20.9
  244. 244 Sieve 0.8B 20.8
  245. 245 Strands Decider 2B 20.7
  246. 246 EXAONE-4.0-1.2B-JEV v0.3 20.6
  247. 247 tasksource Decider nano v5 (150M) 19.9
  248. 248 Crivo-68M 19.5
  249. 249 Amber 68M 19.2
  250. 250 JEV Ettin 150M 18.5
  251. 251 vLLM-SR Decision 2.0 Eos 0.8B 18.4
  252. 252 Crivo-150M v2 18.3
  253. 253 Bekko System One v0 400M 18.0
  254. 254 Mola-150M 17.8
  255. 255 vLLM-SR d3-edge (0.6B) 17.7
  256. 256 Mola-68M 17.6
  257. 257 JEV FLAN-T5 Large public v2 (342M) 17.3
  258. 258 DecisionTune 1.0 17.2
  259. 259 Bosun v3.1 1.7B 16.5
  260. 260 vLLM-SR Decision 2.0 Kai 0.6B 16.4
  261. 261 Dinah-1 16.4
  262. 262 Kev 0.8B r15 16.3
  263. 263 Amber 32M 16.0
  264. 264 PICK-150M v5 15.6
  265. 265 JPT-0.8B 15.5
  266. 266 JEV Ettin 400M 15.2
  267. 267 Sifr 0.8B v3.1 14.5
  268. 268 Crivo-32M v2 14.2
  269. 269 Prumo 14.0
  270. 270 Kodiak v0.2 1B (accuracy) 13.9
  271. 271 Intern-Decision-0.8B 12.7
  272. 272 Jev-Style 0.8B v3 12.5
  273. 273 Tev1-0.8B-experimental 12.5
  274. 274 Bekko System One v0 68M 12.4
  275. 275 Amber 17M 12.0
  276. 276 LiquidAI d1-omni-600M 11.8
  277. 277 Crivo-17M 11.7
  278. 278 Von 1.3 11.7
  279. 279 Bosun v3.1 0.6B 11.0
  280. 280 vera-spark 11.0
  281. 281 Lavoir 10.8
  282. 282 GLiNER2.5-Decide 10.1
  283. 283 Laya 9.2
  284. 284 JevK5 Lite 8.8
  285. 285 LFM2.5-2.6B-RLCD 8.2
  286. 286 Qwen-2.5-1B-RLCD 7.8
  287. 287 Bekko System One v0 17M 7.7
  288. 288 EmbeddingGemma 2 System-One 7.4
  289. 289 CLM-v0.1-8B 6.5
  290. 290 jeff 6.2
  291. 291 LFM2.5-350M-RLCD 5.9
  292. 292 system-one-gemma 5.2
  293. 293 DM-JEPA 1.1 5.0
  294. 294 Julia 1 4.8
  295. 295 Nirnaya 0.4B 3.5
  296. 296 Zev 3.4
  297. 297 Verdict 2.0

Source: Official Decision Index leaderboard

#EntryScoreSettingDateSource
1 karpo-decision-31B 73.9 Official leaderboard
2 GLiDE 27B 73.4 Official leaderboard
3 REAX 1.3 72.6 Official leaderboard
4 Logits Sigma-1 Duo 72.3 Official leaderboard
5 Copernicus-31B 72.1 Official leaderboard
6 KnowLine 31B Gen1.5 71.9 Official leaderboard
7 Copernicus-57B 71.8 Official leaderboard
8 FlexDM-LoRA 2.1 · Gemma-4-31B 71.6 Official leaderboard
9 Sieve 31B v3 71.1 Official leaderboard
10 OneJev1.6g-31B 70.9 Official leaderboard
11 Turion Jev 31B 70.5 Official leaderboard
12 eshbar-lulilla (Gemma 4 31B) 70.3 Official leaderboard
13 Jeffan-27B-1211 70.3 Official leaderboard
14 vLLM-SR d3-ultra (Decision 3.0, 27B) 70.2 Official leaderboard
15 Bespoke Nimble v4.1 Max (27B) 70.0 Official leaderboard
16 H2O Lightning 31B v1.3.0 69.9 Official leaderboard
17 Jeffan-31b-1710 69.8 Official leaderboard
18 Logits Sigma-1 69.5 Official leaderboard
19 liquamen (Gemma 4 31B) 69.1 Official leaderboard
20 Cerno 27B v1.1 69.0 Official leaderboard
21 WeMM-JEV-27B 68.5 Official leaderboard
22 baselabs mammoth-1 (27B) 68.4 Official leaderboard
23 Piccolo 27B 68.4 Official leaderboard
24 Ainvest-Decision-27B v1.0 68.2 Official leaderboard
25 oenogarum (Gemma 4 31B) 68.1 Official leaderboard
26 Perplexity Decider v1.2 (27B) 68.0 Official leaderboard
27 AutoTrust-JEV-27B v1.1 68.0 Official leaderboard
28 StartLux-Decision v1.1 27B 67.9 Official leaderboard
29 Perplexity Decider v1.3 (27B) 67.7 Official leaderboard
30 MISA AI 2.0 Decision Plus · oak 1.1 (31B) 67.6 Official leaderboard
31 AJev lm6q (Gemma 4 31B) 67.6 Official leaderboard
32 Gemma 4 31B IT 67.6 FlexDM Apr 2026 Official leaderboard
33 Chon-1.1-31B 67.5 Official leaderboard
34 decy-preview (31B) 67.5 Official leaderboard
35 Lychee-27B 67.3 Official leaderboard
36 vLLM-SR d3-turbo v1.1 (27B) 67.3 Official leaderboard
37 vLLM-SR d3 (27B) 67.2 Official leaderboard
38 allec-oavg (Gemma 4 31B) 67.0 Official leaderboard
39 VisionJEV 1.0.0 (27B) 66.9 Official leaderboard
40 Aleria Judge 1.1.1 (Gemma 4 31B) 66.7 Official leaderboard
41 Nace.AI Drex Max v1 (27B) 66.6 Official leaderboard
42 AnyJev Tacitus 31B 66.5 Official leaderboard
43 vera-large-2o 66.3 Official leaderboard
44 Decider 31B (NVFP4) 66.0 Official leaderboard
45 KnowLine 35B-A3B Gen1.5 65.8 Official leaderboard
46 OneShot 27B 65.7 Official leaderboard
47 Valen-Decision 1.5 27B 65.2 Official leaderboard
48 Bobcat 1.3 (31B) 64.8 Official leaderboard
49 Amos 1 Large (27B) 64.8 Official leaderboard
50 Matilda Decision v1 64.8 Official leaderboard
51 KnowLine 26B-A4B Gen2.5 64.6 Official leaderboard
52 frontier-infra Jebadiah 27B v2.1 (t1e) 64.4 Official leaderboard
53 Darwin-27B-RSI-Decision 64.2 Official leaderboard
54 deck31b (Gemma 4 31B) 63.0 Official leaderboard
55 decisio 0.6.0 (Gemma 4 31B) 62.7 Official leaderboard
56 GLiDE 9B 62.5 Official leaderboard
57 Torchcast Decision 27B 61.5 Official leaderboard
58 Jeffan-9B-1211 61.5 Official leaderboard
59 OneJev1.6-27B 61.5 Official leaderboard
60 Qwen Decision 27B noncausal LoRA 61.1 Official leaderboard
61 Quyet-1.0-Large 60.8 Official leaderboard
62 garum (Gemma 4 31B) 60.3 Official leaderboard
63 baselabs elephant-1 (9B) 60.3 Official leaderboard
64 Standard One 35B SH (v1) 60.1 Official leaderboard
65 NTele-Hybrid-27B-v1 60.1 Official leaderboard
66 Lumma-fev-27B 60.0 Official leaderboard
67 Sieve 27B v2 59.9 Official leaderboard
68 mogan-decision-31B-it 59.8 Official leaderboard
69 linne-fern 1.0 (27B) 59.6 Official leaderboard
70 Kev 27B 59.6 Official leaderboard
71 Xor 26B-A4B v1.2 59.4 Official leaderboard
72 Jev 1.13 59.2 Official leaderboard
73 Adjudex-9B v5 59.2 Official leaderboard
74 span-02 (Gemma 4 26B-A4B) 59.1 Official leaderboard
75 Cerno 9B v1.2 59.0 Official leaderboard
76 garas (Gemma 4 31B) 59.0 Official leaderboard
77 Shisa DE-2.1 (repeat) 58.8 Official leaderboard
78 Eikos-27B 58.8 Official leaderboard
79 reflex Qwen3.8-27B-FP8 (pmscr) 58.8 Official leaderboard
80 RSI-Jev v6.1 VL 27B (effort auto) 58.6 Official leaderboard
81 KnowLine 9B Gen1 58.4 Official leaderboard
82 H2O Lightning Bolt 74B v1.1 58.4 Official leaderboard
83 StartLux-Decision v1.1 9B 58.1 Official leaderboard
84 vLLM-SR Decision 2.0 Vega 27B 58.0 Official leaderboard
85 JEV-35B 57.6 Official leaderboard
86 JEV-27B 57.4 Official leaderboard
87 Piccolo 9B 57.3 Official leaderboard
88 Qwen 3.8 27B 57.0 simple-jev Aug 2026 Official leaderboard
89 Tern-35B-A3B 56.6 Official leaderboard
90 Surogate Rune 26B-A4B v3 56.4 Official leaderboard
91 SynACK Decide 26B-A4B v7 56.2 Official leaderboard
92 KnowLine 4B Gen5.5 56.0 Official leaderboard
93 Blink v0.3 26B-A4B (NVFP4) 55.9 Official leaderboard
94 Valen-Decision 1.5 9B 55.9 Official leaderboard
95 JADE 55.6 Official leaderboard
96 Dexy-Decision-9B-V2 55.4 Official leaderboard
97 Wald-26B-A4B 55.4 Official leaderboard
98 AJev lora1 (Gemma 4 26B-A4B) 55.3 Official leaderboard
99 Perplexity Decider Flash v1.3 (4B) 55.1 Official leaderboard
100 Cloudflare clef 54.9 Official leaderboard
101 StartLux-Decision v1.1 4B 54.8 Official leaderboard
102 Qwen3.6 27B 54.7 Decider chat Apr 2026 Official leaderboard
103 Perplexity Decider Flash v1.2 (4B) 54.3 Official leaderboard
104 vLLM-SR d3-flash (9B) 54.3 Official leaderboard
105 JPT-35B-A3B 54.1 Official leaderboard
106 chinese-decision-pointer-4b 54.0 Official leaderboard
107 Cerno 4B v1.2 53.9 Official leaderboard
108 Xor 1.2 53.6 Official leaderboard
109 GEV-26B-Decide 53.5 Official leaderboard
110 OneJev1.6-9B 53.0 Official leaderboard
111 vera-medium-omni 53.0 Official leaderboard
112 Senova-4B v4 52.8 Official leaderboard
113 Nace.AI Drex v2.0 (35B-A3B) 52.8 Official leaderboard
114 muria (Gemma 4 12B) 52.5 Official leaderboard
115 Jeffan-4B-2010 52.2 Official leaderboard
116 Verdin-4B v1.1 52.0 Official leaderboard
117 LatentMesh Blitz 4B v1 51.9 Official leaderboard
118 Saif 4B v2 51.7 Official leaderboard
119 vev-4b-v2 51.1 Official leaderboard
120 Bespoke Nimble 9B v3 51.1 Official leaderboard
121 frontier-infra Jebadiah 9B v2.1 (t1e) 50.9 Official leaderboard
122 decisio 0.4.0 (Qwen3.6-35B-A3B) 50.8 Official leaderboard
123 Piccolo 4B 50.0 Official leaderboard
124 vLLM-SR d3-mini (4B) 50.0 Official leaderboard
125 RSI-Jev v7.0 VL 4B 49.9 Official leaderboard
126 Valen-Decision 1.5 4B 49.9 Official leaderboard
127 Jegue 0.2 (9B) 49.8 Official leaderboard
128 RSI-Jev v7.0 VL 3B 49.7 Official leaderboard
129 JoshuaSP diffusiongemma (open-jev) 49.5 Official leaderboard
130 AJev lora7 (Gemma 4 12B) 48.9 Official leaderboard
131 OneJev1.6-4B 48.9 Official leaderboard
132 decisio 0.4.0 (Gemma 4 12B) 48.8 Official leaderboard
133 BaseDecision-Dyno (4B) 48.5 Official leaderboard
134 H2O Lightning E4B v1.0.1 48.0 Official leaderboard
135 Mica v0.5 4B 47.7 Official leaderboard
136 PurpleMIST-Flash-1.0 47.5 Official leaderboard
137 Sieve 9B v2 47.0 Official leaderboard
138 Decider 35B-A3B 46.6 Official leaderboard
139 Winnow-12B 46.6 Official leaderboard
140 Seb-9B v2 46.3 Official leaderboard
141 PurpleMIST-Mini-2.0 46.2 Official leaderboard
142 vLLM-SR Decision 2.0 Lux 9B 46.0 Official leaderboard
143 Nace.AI Drex v1.5 45.8 Official leaderboard
144 Hopper 12B (trained) 45.8 Official leaderboard
145 Trio-Spark DI (9B) 45.6 Official leaderboard
146 Jevora 9B v1.6 45.6 Official leaderboard
147 CoCo-Decision-4B 2.4 45.5 Official leaderboard
148 Cloudflare clef-omni (30B-A3B) 45.0 Official leaderboard
149 JevK5 9B v0.3.3 44.9 Official leaderboard
150 Cloudflare clef-flash 44.4 Official leaderboard
151 Jevfire 44.4 Official leaderboard
152 jiwo 4B v2 44.1 Official leaderboard
153 JPT-9B 44.1 Official leaderboard
154 ezjev 4B s2 43.9 Official leaderboard
155 Jev-Omni 43.8 Official leaderboard
156 GLiDE 2B 43.7 Official leaderboard
157 Solomon v1.1 43.2 Official leaderboard
158 Standard One 8B SH (v4) 42.9 Official leaderboard
159 Amber Hybrid 4B 42.7 Official leaderboard
160 JEV-9B 42.1 Official leaderboard
161 Bespoke Nimble 9B v2 42.1 Official leaderboard
162 Sieve 9B 41.7 Official leaderboard
163 Sieve 4B 41.0 Official leaderboard
164 Kev 9B v2 40.9 Official leaderboard
165 Nari Decision 4B v3.2 40.9 Official leaderboard
166 Aplomb 1 40.8 Official leaderboard
167 vLLM-SR Decision 2.0 Nox 4B 40.8 Official leaderboard
168 Hopper (G) 1.2 40.2 Official leaderboard
169 StartLux-Decision v1.1 2B 40.1 Official leaderboard
170 FlyMy.AI Decision 4B v1.6 39.8 Official leaderboard
171 Standard One 3B SH (v4) 39.8 Official leaderboard
172 JPT-4B 39.6 Official leaderboard
173 Amber 4B 39.5 Official leaderboard
174 Intern-Decision-4B 38.7 Official leaderboard
175 djev 38.1 Official leaderboard
176 KnowLine 2B Gen1 38.1 Official leaderboard
177 Valen-Decision 1.5 2B 38.0 Official leaderboard
178 razorback16 openjev diffusiongemma (NVFP4, vLLM) 37.6 Official leaderboard
179 NeoHorse-Jev-4B 37.3 Official leaderboard
180 JevK5 v0.3 36.6 Official leaderboard
181 GLiDE 0.8B 36.4 Official leaderboard
182 Decider 4B 36.3 Official leaderboard
183 baselabs joey-1 (2B) 35.9 Official leaderboard
184 lev 35.5 Official leaderboard
185 Lev1 3.7B 35.4 Official leaderboard
186 Jev-Style 2B v4.2 35.1 Official leaderboard
187 Kev 4B r10 34.9 Official leaderboard
188 PKU-KCL-Meteor-2B 34.8 Official leaderboard
189 Sieve 2B v3 34.5 Official leaderboard
190 LatentMesh Blitz 2B v1 34.4 Official leaderboard
191 Saracura PT-BR 4B 34.4 Official leaderboard
192 LiquidAI d1-2B 34.4 Official leaderboard
193 Cam1 34.4 Official leaderboard
194 vLLM-SR d3-nano (2B) 34.3 Official leaderboard
195 mmastrac diffusiongemma (vLLM PR 57250) 34.3 Official leaderboard
196 Kas 4B 34.0 Official leaderboard
197 GLiDE 0.4B 33.8 Official leaderboard
198 Tev1-4B-experimental 33.7 Official leaderboard
199 LiquidAI d1-Omni-2B 33.6 Official leaderboard
200 Adjudex-0.8B v5 32.8 Official leaderboard
201 Winnow-E4B 32.7 Official leaderboard
202 Aurelis-2B 32.5 Official leaderboard
203 GLiClass v4.1 edge (0.4B) 32.5 Official leaderboard
204 Jevstral 8B 32.2 Official leaderboard
205 Metask-Jev-4B 32.2 Official leaderboard
206 Ryotide Qwen v0.3.1 32.0 Official leaderboard
207 Intelif Qwen3-4B 31.3 Official leaderboard
208 LiquidAI d1-3B 31.1 Official leaderboard
209 open-jev (pngwn) 31.1 Official leaderboard
210 PKU-KCL-Meteor-0.8B 30.9 Official leaderboard
211 vera-core-2 30.6 Official leaderboard
212 Jobe 30.4 Official leaderboard
213 OneJev1.6-0.8B 29.6 Official leaderboard
214 crystal-0.8b 29.4 Official leaderboard
215 MoDe-L 29.4 Official leaderboard
216 vLLM-SR vela3-fast (Vela 3.0, 0.4B) 29.3 Official leaderboard
217 Crivo-1B 29.3 Official leaderboard
218 KnowLine 0.8B Gen1 29.2 Official leaderboard
219 Gevva E4B 29.2 Official leaderboard
220 StartLux-Decision v1.1 0.8B 28.9 Official leaderboard
221 openvons 28.8 Official leaderboard
222 Amber 0.4B v2 27.8 Official leaderboard
223 Valen-Decision 1.5 0.8B 27.6 Official leaderboard
224 SemIf 26.7 Official leaderboard
225 Saina Helm 2 Omni (0.8B) 26.5 Official leaderboard
226 MoDe-M 26.5 Official leaderboard
227 Saif 0.8B Lightning v0.3 26.4 Official leaderboard
228 vLLM-SR d3-lite (0.8B) 26.2 Official leaderboard
229 jiwo 0.8B v2 25.8 Official leaderboard
230 Saina Helm 2.0 v2 25.3 Official leaderboard
231 this-that 1.2 24.9 Official leaderboard
232 Amber 150M 24.7 Official leaderboard
233 MoDe-S 24.5 Official leaderboard
234 Gevva E2B 24.3 Official leaderboard
235 Crivo-400M 23.7 Official leaderboard
236 Decider 2B 23.6 Official leaderboard
237 vLLM-SR Decision 2.0 Sol 2B 23.6 Official leaderboard
238 jdi-ptr14-0.8b 23.6 Official leaderboard
239 mini-jev 22.8 Official leaderboard
240 vLLM-SR vela3-fast-vision (Vela 3.0, 0.25B) 22.1 Official leaderboard
241 Sieve 2B 21.9 Official leaderboard
242 wyat-1-800m 21.5 Official leaderboard
243 Intern-Decision-2B 20.9 Official leaderboard
244 Sieve 0.8B 20.8 Official leaderboard
245 Strands Decider 2B 20.7 Official leaderboard
246 EXAONE-4.0-1.2B-JEV v0.3 20.6 Official leaderboard
247 tasksource Decider nano v5 (150M) 19.9 Official leaderboard
248 Crivo-68M 19.5 Official leaderboard
249 Amber 68M 19.2 Official leaderboard
250 JEV Ettin 150M 18.5 Official leaderboard
251 vLLM-SR Decision 2.0 Eos 0.8B 18.4 Official leaderboard
252 Crivo-150M v2 18.3 Official leaderboard
253 Bekko System One v0 400M 18.0 Official leaderboard
254 Mola-150M 17.8 Official leaderboard
255 vLLM-SR d3-edge (0.6B) 17.7 Official leaderboard
256 Mola-68M 17.6 Official leaderboard
257 JEV FLAN-T5 Large public v2 (342M) 17.3 Official leaderboard
258 DecisionTune 1.0 17.2 Official leaderboard
259 Bosun v3.1 1.7B 16.5 Official leaderboard
260 vLLM-SR Decision 2.0 Kai 0.6B 16.4 Official leaderboard
261 Dinah-1 16.4 Official leaderboard
262 Kev 0.8B r15 16.3 Official leaderboard
263 Amber 32M 16.0 Official leaderboard
264 PICK-150M v5 15.6 Official leaderboard
265 JPT-0.8B 15.5 Official leaderboard
266 JEV Ettin 400M 15.2 Official leaderboard
267 Sifr 0.8B v3.1 14.5 Official leaderboard
268 Crivo-32M v2 14.2 Official leaderboard
269 Prumo 14.0 Official leaderboard
270 Kodiak v0.2 1B (accuracy) 13.9 Official leaderboard
271 Intern-Decision-0.8B 12.7 Official leaderboard
272 Jev-Style 0.8B v3 12.5 Official leaderboard
273 Tev1-0.8B-experimental 12.5 Official leaderboard
274 Bekko System One v0 68M 12.4 Official leaderboard
275 Amber 17M 12.0 Official leaderboard
276 LiquidAI d1-omni-600M 11.8 Official leaderboard
277 Crivo-17M 11.7 Official leaderboard
278 Von 1.3 11.7 Official leaderboard
279 Bosun v3.1 0.6B 11.0 Official leaderboard
280 vera-spark 11.0 Official leaderboard
281 Lavoir 10.8 Official leaderboard
282 GLiNER2.5-Decide 10.1 Official leaderboard
283 Laya 9.2 Official leaderboard
284 JevK5 Lite 8.8 Official leaderboard
285 LFM2.5-2.6B-RLCD 8.2 Official leaderboard
286 Qwen-2.5-1B-RLCD 7.8 Official leaderboard
287 Bekko System One v0 17M 7.7 Official leaderboard
288 EmbeddingGemma 2 System-One 7.4 Official leaderboard
289 CLM-v0.1-8B 6.5 Official leaderboard
290 jeff 6.2 Official leaderboard
291 LFM2.5-350M-RLCD 5.9 Official leaderboard
292 system-one-gemma 5.2 Official leaderboard
293 DM-JEPA 1.1 5.0 Official leaderboard
294 Julia 1 4.8 Official leaderboard
295 Nirnaya 0.4B 3.5 Official leaderboard
296 Zev 3.4 Official leaderboard
297 Verdict 2.0 Official leaderboard

Median latency vs Decision Index

Median time per request on one RTX PRO 6000, on a log scale; for hosted APIs it is an HTTP round trip, so not a controlled comparison. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0 20 40 60 80 0.00010.0010.010.11 Median latency (s), log scale, lower is better Decision Index (Full) Gemma 4 31B IT: Decision Index (Full) 61.6, median latency 0.11 s Gemma 4 31B IT: Decision Index (Full) 63.6, median latency 0.1 s Gemma 4 31B IT: Decision Index (Full) 64.6, median latency 0.1 s Gemma 4 31B IT: Decision Index (Full) 67.6, median latency 0.11 s Qwen3.6 27B: Decision Index (Full) 54.7, median latency 0.08 s Qwen 3.8 27B: Decision Index (Full) 57, median latency 0.37 s GLiDE 27B: Decision Index (Full) 73.4, median latency 0.12 s. Not in our model list REAX 1.3: Decision Index (Full) 72.6, median latency 0.25 s. Not in our model list Logits Sigma-1 Duo: Decision Index (Full) 72.3, median latency 0.37 s. Not in our model list Copernicus-31B: Decision Index (Full) 72.1, median latency 0.15 s. Not in our model list KnowLine 31B Gen1.5: Decision Index (Full) 71.9, median latency 0.2 s. Not in our model list Copernicus-57B: Decision Index (Full) 71.8, median latency 0.25 s. Not in our model list FlexDM-LoRA 2.1 · Gemma-4-31B: Decision Index (Full) 71.6, median latency 0.15 s. Not in our model list Sieve 31B v3: Decision Index (Full) 71.1, median latency 0.23 s. Not in our model list OneJev1.6g-31B: Decision Index (Full) 70.9, median latency 0.09 s. Not in our model list Turion Jev 31B: Decision Index (Full) 70.5, median latency 0.14 s. Not in our model list eshbar-lulilla (Gemma 4 31B): Decision Index (Full) 70.3, median latency 0.34 s. Not in our model list Jeffan-27B-1211: Decision Index (Full) 70.3, median latency 0.1 s. Not in our model list vLLM-SR d3-ultra (Decision 3.0, 27B): Decision Index (Full) 70.2, median latency 0.2 s. Not in our model list Bespoke Nimble v4.1 Max (27B): Decision Index (Full) 70, median latency 0.11 s. Not in our model list H2O Lightning 31B v1.3.0: Decision Index (Full) 69.9, median latency 0.14 s. Not in our model list Jeffan-31b-1710: Decision Index (Full) 69.8, median latency 0.12 s. Not in our model list Logits Sigma-1: Decision Index (Full) 69.5, median latency 0.12 s. Not in our model list liquamen (Gemma 4 31B): Decision Index (Full) 69.1, median latency 0.32 s. Not in our model list WeMM-JEV-27B: Decision Index (Full) 68.5, median latency 0.1 s. Not in our model list baselabs mammoth-1 (27B): Decision Index (Full) 68.4, median latency 0.22 s. Not in our model list Piccolo 27B: Decision Index (Full) 68.4, median latency 0.1 s. Not in our model list Ainvest-Decision-27B v1.0: Decision Index (Full) 68.2, median latency 0.1 s. Not in our model list oenogarum (Gemma 4 31B): Decision Index (Full) 68.1, median latency 0.26 s. Not in our model list Perplexity Decider v1.2 (27B): Decision Index (Full) 68, median latency 0.14 s. Not in our model list Perplexity Decider v1.3 (27B): Decision Index (Full) 67.7, median latency 0.14 s. Not in our model list MISA AI 2.0 Decision Plus · oak 1.1 (31B): Decision Index (Full) 67.6, median latency 0.11 s. Not in our model list AJev lm6q (Gemma 4 31B): Decision Index (Full) 67.6, median latency 0.14 s. Not in our model list Chon-1.1-31B: Decision Index (Full) 67.5, median latency 0.17 s. Not in our model list decy-preview (31B): Decision Index (Full) 67.5, median latency 0.12 s. Not in our model list Lychee-27B: Decision Index (Full) 67.3, median latency 0.1 s. Not in our model list vLLM-SR d3-turbo v1.1 (27B): Decision Index (Full) 67.3, median latency 0.19 s. Not in our model list vLLM-SR d3 (27B): Decision Index (Full) 67.2, median latency 0.1 s. Not in our model list allec-oavg (Gemma 4 31B): Decision Index (Full) 67, median latency 0.26 s. Not in our model list VisionJEV 1.0.0 (27B): Decision Index (Full) 66.9, median latency 0.09 s. Not in our model list Aleria Judge 1.1.1 (Gemma 4 31B): Decision Index (Full) 66.7, median latency 0.12 s. Not in our model list Nace.AI Drex Max v1 (27B): Decision Index (Full) 66.6, median latency 0.1 s. Not in our model list AnyJev Tacitus 31B: Decision Index (Full) 66.5, median latency 0.1 s. Not in our model list vera-large-2o: Decision Index (Full) 66.3, median latency 0.15 s. Not in our model list OneShot 27B: Decision Index (Full) 65.7, median latency 0.13 s. Not in our model list Valen-Decision 1.5 27B: Decision Index (Full) 65.2, median latency 0.11 s. Not in our model list Bobcat 1.3 (31B): Decision Index (Full) 64.8, median latency 0.06 s. Not in our model list Amos 1 Large (27B): Decision Index (Full) 64.8, median latency 0.15 s. Not in our model list Matilda Decision v1: Decision Index (Full) 64.8, median latency 0.09 s. Not in our model list KnowLine 26B-A4B Gen2.5: Decision Index (Full) 64.6, median latency 0.07 s. Not in our model list frontier-infra Jebadiah 27B v2.1 (t1e): Decision Index (Full) 64.4, median latency 0.3 s. Not in our model list Darwin-27B-RSI-Decision: Decision Index (Full) 64.2, median latency 0.12 s. Not in our model list deck31b (Gemma 4 31B): Decision Index (Full) 63, median latency 0.22 s. Not in our model list decisio 0.6.0 (Gemma 4 31B): Decision Index (Full) 62.7, median latency 0.05 s. Not in our model list GLiDE 9B: Decision Index (Full) 62.5, median latency 0.09 s. Not in our model list Torchcast Decision 27B: Decision Index (Full) 61.5, median latency 0.1 s. Not in our model list Jeffan-9B-1211: Decision Index (Full) 61.5, median latency 0.08 s. Not in our model list OneJev1.6-27B: Decision Index (Full) 61.5, median latency 0.11 s. Not in our model list Qwen Decision 27B noncausal LoRA: Decision Index (Full) 61.1, median latency 0.13 s. Not in our model list Quyet-1.0-Large: Decision Index (Full) 60.8, median latency 0.11 s. Not in our model list garum (Gemma 4 31B): Decision Index (Full) 60.3, median latency 0.12 s. Not in our model list baselabs elephant-1 (9B): Decision Index (Full) 60.3, median latency 0.11 s. Not in our model list Standard One 35B SH (v1): Decision Index (Full) 60.1, median latency 0.32 s. Not in our model list NTele-Hybrid-27B-v1: Decision Index (Full) 60.1, median latency 0.08 s. Not in our model list Lumma-fev-27B: Decision Index (Full) 60, median latency 0.1 s. Not in our model list Sieve 27B v2: Decision Index (Full) 59.9, median latency 0.11 s. Not in our model list mogan-decision-31B-it: Decision Index (Full) 59.8, median latency 0.11 s. Not in our model list linne-fern 1.0 (27B): Decision Index (Full) 59.6, median latency 0.11 s. Not in our model list Kev 27B: Decision Index (Full) 59.6, median latency 0.14 s. Not in our model list Xor 26B-A4B v1.2: Decision Index (Full) 59.4, median latency 0.11 s. Not in our model list Jev 1.13: Decision Index (Full) 59.2, median latency 0.52 s. Not in our model list Adjudex-9B v5: Decision Index (Full) 59.2, median latency 0.07 s. Not in our model list Cerno 9B v1.2: Decision Index (Full) 59, median latency 0.07 s. Not in our model list garas (Gemma 4 31B): Decision Index (Full) 59, median latency 0.12 s. Not in our model list Shisa DE-2.1 (repeat): Decision Index (Full) 58.8, median latency 0.05 s. Not in our model list Eikos-27B: Decision Index (Full) 58.8, median latency 0.13 s. Not in our model list reflex Qwen3.8-27B-FP8 (pmscr): Decision Index (Full) 58.8, median latency 0.15 s. Not in our model list RSI-Jev v6.1 VL 27B (effort auto): Decision Index (Full) 58.6, median latency 0.08 s. Not in our model list KnowLine 9B Gen1: Decision Index (Full) 58.4, median latency 0.13 s. Not in our model list H2O Lightning Bolt 74B v1.1: Decision Index (Full) 58.4, median latency 0.29 s. Not in our model list StartLux-Decision v1.1 9B: Decision Index (Full) 58.1, median latency 0.04 s. Not in our model list vLLM-SR Decision 2.0 Vega 27B: Decision Index (Full) 58, median latency 0.12 s. Not in our model list JEV-35B: Decision Index (Full) 57.6, median latency 0.05 s. Not in our model list JEV-27B: Decision Index (Full) 57.4, median latency 0.1 s. Not in our model list Piccolo 9B: Decision Index (Full) 57.3, median latency 0.08 s. Not in our model list Tern-35B-A3B: Decision Index (Full) 56.6, median latency 0.06 s. Not in our model list Surogate Rune 26B-A4B v3: Decision Index (Full) 56.4, median latency 0.12 s. Not in our model list SynACK Decide 26B-A4B v7: Decision Index (Full) 56.2, median latency 0.13 s. Not in our model list KnowLine 4B Gen5.5: Decision Index (Full) 56, median latency 0.04 s. Not in our model list Valen-Decision 1.5 9B: Decision Index (Full) 55.9, median latency 0.12 s. Not in our model list JADE: Decision Index (Full) 55.6, median latency 0.13 s. Not in our model list Dexy-Decision-9B-V2: Decision Index (Full) 55.4, median latency 0.03 s. Not in our model list Wald-26B-A4B: Decision Index (Full) 55.4, median latency 0.09 s. Not in our model list AJev lora1 (Gemma 4 26B-A4B): Decision Index (Full) 55.3, median latency 0.09 s. Not in our model list Perplexity Decider Flash v1.3 (4B): Decision Index (Full) 55.1, median latency 0.09 s. Not in our model list Cloudflare clef: Decision Index (Full) 54.9, median latency 0.1 s. Not in our model list StartLux-Decision v1.1 4B: Decision Index (Full) 54.8, median latency 0.04 s. Not in our model list Perplexity Decider Flash v1.2 (4B): Decision Index (Full) 54.3, median latency 0.07 s. Not in our model list vLLM-SR d3-flash (9B): Decision Index (Full) 54.3, median latency 0.05 s. Not in our model list JPT-35B-A3B: Decision Index (Full) 54.1, median latency 0.25 s. Not in our model list chinese-decision-pointer-4b: Decision Index (Full) 54, median latency 0.05 s. Not in our model list Cerno 4B v1.2: Decision Index (Full) 53.9, median latency 0.06 s. Not in our model list Xor 1.2: Decision Index (Full) 53.6, median latency 0.15 s. Not in our model list GEV-26B-Decide: Decision Index (Full) 53.5, median latency 0.13 s. Not in our model list OneJev1.6-9B: Decision Index (Full) 53, median latency 0.04 s. Not in our model list vera-medium-omni: Decision Index (Full) 53, median latency 0.09 s. Not in our model list Senova-4B v4: Decision Index (Full) 52.8, median latency 0.08 s. Not in our model list Nace.AI Drex v2.0 (35B-A3B): Decision Index (Full) 52.8, median latency 0.28 s. Not in our model list muria (Gemma 4 12B): Decision Index (Full) 52.5, median latency 0.05 s. Not in our model list Jeffan-4B-2010: Decision Index (Full) 52.2, median latency 0.05 s. Not in our model list Verdin-4B v1.1: Decision Index (Full) 52, median latency 0.2 s. Not in our model list LatentMesh Blitz 4B v1: Decision Index (Full) 51.9, median latency 0.07 s. Not in our model list Saif 4B v2: Decision Index (Full) 51.7, median latency 0.07 s. Not in our model list vev-4b-v2: Decision Index (Full) 51.1, median latency 0.06 s. Not in our model list Bespoke Nimble 9B v3: Decision Index (Full) 51.1, median latency 0.07 s. Not in our model list frontier-infra Jebadiah 9B v2.1 (t1e): Decision Index (Full) 50.9, median latency 0.13 s. Not in our model list decisio 0.4.0 (Qwen3.6-35B-A3B): Decision Index (Full) 50.8, median latency 0.07 s. Not in our model list Piccolo 4B: Decision Index (Full) 50, median latency 0.05 s. Not in our model list vLLM-SR d3-mini (4B): Decision Index (Full) 50, median latency 0.05 s. Not in our model list RSI-Jev v7.0 VL 4B: Decision Index (Full) 49.9, median latency 0.06 s. Not in our model list Valen-Decision 1.5 4B: Decision Index (Full) 49.9, median latency 0.05 s. Not in our model list Jegue 0.2 (9B): Decision Index (Full) 49.8, median latency 0.09 s. Not in our model list RSI-Jev v7.0 VL 3B: Decision Index (Full) 49.7, median latency 0.1 s. Not in our model list JoshuaSP diffusiongemma (open-jev): Decision Index (Full) 49.5, median latency 0.26 s. Not in our model list AJev lora7 (Gemma 4 12B): Decision Index (Full) 48.9, median latency 0.07 s. Not in our model list OneJev1.6-4B: Decision Index (Full) 48.9, median latency 0.03 s. Not in our model list decisio 0.4.0 (Gemma 4 12B): Decision Index (Full) 48.8, median latency 0.04 s. Not in our model list BaseDecision-Dyno (4B): Decision Index (Full) 48.5, median latency 0.14 s. Not in our model list PurpleMIST-Flash-1.0: Decision Index (Full) 47.5, median latency 0.05 s. Not in our model list Sieve 9B v2: Decision Index (Full) 47, median latency 0.05 s. Not in our model list Decider 35B-A3B: Decision Index (Full) 46.6, median latency 0.1 s. Not in our model list Winnow-12B: Decision Index (Full) 46.6, median latency 0.07 s. Not in our model list Seb-9B v2: Decision Index (Full) 46.3, median latency 0.09 s. Not in our model list PurpleMIST-Mini-2.0: Decision Index (Full) 46.2, median latency 0.06 s. Not in our model list vLLM-SR Decision 2.0 Lux 9B: Decision Index (Full) 46, median latency 0.04 s. Not in our model list Nace.AI Drex v1.5: Decision Index (Full) 45.8, median latency 0.05 s. Not in our model list Hopper 12B (trained): Decision Index (Full) 45.8, median latency 0.29 s. Not in our model list Trio-Spark DI (9B): Decision Index (Full) 45.6, median latency 0.21 s. Not in our model list Jevora 9B v1.6: Decision Index (Full) 45.6, median latency 0.12 s. Not in our model list CoCo-Decision-4B 2.4: Decision Index (Full) 45.5, median latency 0.06 s. Not in our model list Cloudflare clef-omni (30B-A3B): Decision Index (Full) 45, median latency 0.19 s. Not in our model list JevK5 9B v0.3.3: Decision Index (Full) 44.9, median latency 0.04 s. Not in our model list Cloudflare clef-flash: Decision Index (Full) 44.4, median latency 0.05 s. Not in our model list Jevfire: Decision Index (Full) 44.4, median latency 0.09 s. Not in our model list jiwo 4B v2: Decision Index (Full) 44.1, median latency 0.05 s. Not in our model list JPT-9B: Decision Index (Full) 44.1, median latency 0.15 s. Not in our model list ezjev 4B s2: Decision Index (Full) 43.9, median latency 0.04 s. Not in our model list Jev-Omni: Decision Index (Full) 43.8, median latency 0.05 s. Not in our model list GLiDE 2B: Decision Index (Full) 43.7, median latency 0.04 s. Not in our model list Solomon v1.1: Decision Index (Full) 43.2, median latency 0.22 s. Not in our model list Standard One 8B SH (v4): Decision Index (Full) 42.9, median latency 0.09 s. Not in our model list Amber Hybrid 4B: Decision Index (Full) 42.7, median latency 0.27 s. Not in our model list JEV-9B: Decision Index (Full) 42.1, median latency 0.05 s. Not in our model list Bespoke Nimble 9B v2: Decision Index (Full) 42.1, median latency 0.08 s. Not in our model list Sieve 9B: Decision Index (Full) 41.7, median latency 0.05 s. Not in our model list Sieve 4B: Decision Index (Full) 41, median latency 0.09 s. Not in our model list Kev 9B v2: Decision Index (Full) 40.9, median latency 0.06 s. Not in our model list Nari Decision 4B v3.2: Decision Index (Full) 40.9, median latency 0.02 s. Not in our model list Aplomb 1: Decision Index (Full) 40.8, median latency 0.05 s. Not in our model list vLLM-SR Decision 2.0 Nox 4B: Decision Index (Full) 40.8, median latency 0.03 s. Not in our model list Hopper (G) 1.2: Decision Index (Full) 40.2, median latency 0.02 s. Not in our model list StartLux-Decision v1.1 2B: Decision Index (Full) 40.1, median latency 0.03 s. Not in our model list FlyMy.AI Decision 4B v1.6: Decision Index (Full) 39.8, median latency 0.03 s. Not in our model list Standard One 3B SH (v4): Decision Index (Full) 39.8, median latency 0.08 s. Not in our model list JPT-4B: Decision Index (Full) 39.6, median latency 0.14 s. Not in our model list Amber 4B: Decision Index (Full) 39.5, median latency 0.13 s. Not in our model list Intern-Decision-4B: Decision Index (Full) 38.7, median latency 0.04 s. Not in our model list djev: Decision Index (Full) 38.1, median latency 0.08 s. Not in our model list KnowLine 2B Gen1: Decision Index (Full) 38.1, median latency 0.03 s. Not in our model list Valen-Decision 1.5 2B: Decision Index (Full) 38, median latency 0.04 s. Not in our model list razorback16 openjev diffusiongemma (NVFP4, vLLM): Decision Index (Full) 37.6, median latency 0.04 s. Not in our model list NeoHorse-Jev-4B: Decision Index (Full) 37.3, median latency 0.05 s. Not in our model list JevK5 v0.3: Decision Index (Full) 36.6, median latency 0.03 s. Not in our model list GLiDE 0.8B: Decision Index (Full) 36.4, median latency 0.04 s. Not in our model list baselabs joey-1 (2B): Decision Index (Full) 35.9, median latency 0.08 s. Not in our model list lev: Decision Index (Full) 35.5, median latency 0.07 s. Not in our model list Lev1 3.7B: Decision Index (Full) 35.4, median latency 0.03 s. Not in our model list Jev-Style 2B v4.2: Decision Index (Full) 35.1, median latency 0.03 s. Not in our model list Kev 4B r10: Decision Index (Full) 34.9, median latency 0.06 s. Not in our model list PKU-KCL-Meteor-2B: Decision Index (Full) 34.8, median latency 0.03 s. Not in our model list Sieve 2B v3: Decision Index (Full) 34.5, median latency 0.04 s. Not in our model list LatentMesh Blitz 2B v1: Decision Index (Full) 34.4, median latency 0.05 s. Not in our model list Saracura PT-BR 4B: Decision Index (Full) 34.4, median latency 0.06 s. Not in our model list LiquidAI d1-2B: Decision Index (Full) 34.4, median latency 0.09 s. Not in our model list Cam1: Decision Index (Full) 34.4, median latency 0.25 s. Not in our model list vLLM-SR d3-nano (2B): Decision Index (Full) 34.3, median latency 0.04 s. Not in our model list mmastrac diffusiongemma (vLLM PR 57250): Decision Index (Full) 34.3, median latency 0.13 s. Not in our model list Kas 4B: Decision Index (Full) 34, median latency 0.07 s. Not in our model list GLiDE 0.4B: Decision Index (Full) 33.8, median latency 0.02 s. Not in our model list Tev1-4B-experimental: Decision Index (Full) 33.7, median latency 0.04 s. Not in our model list LiquidAI d1-Omni-2B: Decision Index (Full) 33.6, median latency 0.1 s. Not in our model list Adjudex-0.8B v5: Decision Index (Full) 32.8, median latency 0.05 s. Not in our model list Winnow-E4B: Decision Index (Full) 32.7, median latency 0.04 s. Not in our model list Aurelis-2B: Decision Index (Full) 32.5, median latency 0.07 s. Not in our model list GLiClass v4.1 edge (0.4B): Decision Index (Full) 32.5, median latency 0.02 s. Not in our model list Jevstral 8B: Decision Index (Full) 32.2, median latency 0.03 s. Not in our model list Metask-Jev-4B: Decision Index (Full) 32.2, median latency 0.06 s. Not in our model list Ryotide Qwen v0.3.1: Decision Index (Full) 32, median latency 0.08 s. Not in our model list Intelif Qwen3-4B: Decision Index (Full) 31.3, median latency 0.02 s. Not in our model list LiquidAI d1-3B: Decision Index (Full) 31.1, median latency 0.02 s. Not in our model list open-jev (pngwn): Decision Index (Full) 31.1, median latency 0.11 s. Not in our model list PKU-KCL-Meteor-0.8B: Decision Index (Full) 30.9, median latency 0.04 s. Not in our model list vera-core-2: Decision Index (Full) 30.6, median latency 0.09 s. Not in our model list Jobe: Decision Index (Full) 30.4, median latency 0.05 s. Not in our model list OneJev1.6-0.8B: Decision Index (Full) 29.6, median latency 0.01 s. Not in our model list crystal-0.8b: Decision Index (Full) 29.4, median latency 0.05 s. Not in our model list MoDe-L: Decision Index (Full) 29.4, median latency 0.1 s. Not in our model list vLLM-SR vela3-fast (Vela 3.0, 0.4B): Decision Index (Full) 29.3, median latency 0.05 s. Not in our model list Crivo-1B: Decision Index (Full) 29.3, median latency 0.02 s. Not in our model list KnowLine 0.8B Gen1: Decision Index (Full) 29.2, median latency 0.03 s. Not in our model list Gevva E4B: Decision Index (Full) 29.2, median latency 0.05 s. Not in our model list StartLux-Decision v1.1 0.8B: Decision Index (Full) 28.9, median latency 0.03 s. Not in our model list openvons: Decision Index (Full) 28.8, median latency 0.02 s. Not in our model list Amber 0.4B v2: Decision Index (Full) 27.8, median latency 0.02 s. Not in our model list Valen-Decision 1.5 0.8B: Decision Index (Full) 27.6, median latency 0.04 s. Not in our model list SemIf: Decision Index (Full) 26.7, median latency 0.11 s. Not in our model list Saina Helm 2 Omni (0.8B): Decision Index (Full) 26.5, median latency 0.04 s. Not in our model list MoDe-M: Decision Index (Full) 26.5, median latency 0.12 s. Not in our model list Saif 0.8B Lightning v0.3: Decision Index (Full) 26.4, median latency 0.06 s. Not in our model list vLLM-SR d3-lite (0.8B): Decision Index (Full) 26.2, median latency 0.04 s. Not in our model list jiwo 0.8B v2: Decision Index (Full) 25.8, median latency 0.04 s. Not in our model list Saina Helm 2.0 v2: Decision Index (Full) 25.3, median latency 0.04 s. Not in our model list this-that 1.2: Decision Index (Full) 24.9, median latency 0.04 s. Not in our model list Amber 150M: Decision Index (Full) 24.7, median latency 0.02 s. Not in our model list MoDe-S: Decision Index (Full) 24.5, median latency 0.12 s. Not in our model list Gevva E2B: Decision Index (Full) 24.3, median latency 0.04 s. Not in our model list Crivo-400M: Decision Index (Full) 23.7, median latency 0.02 s. Not in our model list vLLM-SR Decision 2.0 Sol 2B: Decision Index (Full) 23.6, median latency 0.03 s. Not in our model list jdi-ptr14-0.8b: Decision Index (Full) 23.6, median latency 0.07 s. Not in our model list mini-jev: Decision Index (Full) 22.8, median latency 0.07 s. Not in our model list vLLM-SR vela3-fast-vision (Vela 3.0, 0.25B): Decision Index (Full) 22.1, median latency 0.04 s. Not in our model list Sieve 2B: Decision Index (Full) 21.9, median latency 0.06 s. Not in our model list wyat-1-800m: Decision Index (Full) 21.5, median latency 0.11 s. Not in our model list Intern-Decision-2B: Decision Index (Full) 20.9, median latency 0.03 s. Not in our model list Sieve 0.8B: Decision Index (Full) 20.8, median latency 0.05 s. Not in our model list Strands Decider 2B: Decision Index (Full) 20.7, median latency 0.05 s. Not in our model list EXAONE-4.0-1.2B-JEV v0.3: Decision Index (Full) 20.6, median latency 0.15 s. Not in our model list tasksource Decider nano v5 (150M): Decision Index (Full) 19.9, median latency 0.01 s. Not in our model list Crivo-68M: Decision Index (Full) 19.5, median latency 0.01 s. Not in our model list Amber 68M: Decision Index (Full) 19.2, median latency 0.01 s. Not in our model list JEV Ettin 150M: Decision Index (Full) 18.5, median latency 0.08 s. Not in our model list vLLM-SR Decision 2.0 Eos 0.8B: Decision Index (Full) 18.4, median latency 0.03 s. Not in our model list Crivo-150M v2: Decision Index (Full) 18.3, median latency 0.01 s. Not in our model list Bekko System One v0 400M: Decision Index (Full) 18, median latency 0.02 s. Not in our model list Mola-150M: Decision Index (Full) 17.8, median latency 0.01 s. Not in our model list vLLM-SR d3-edge (0.6B): Decision Index (Full) 17.7, median latency 0.02 s. Not in our model list Mola-68M: Decision Index (Full) 17.6, median latency 0.01 s. Not in our model list JEV FLAN-T5 Large public v2 (342M): Decision Index (Full) 17.3, median latency 0.02 s. Not in our model list DecisionTune 1.0: Decision Index (Full) 17.2, median latency 0.01 s. Not in our model list Bosun v3.1 1.7B: Decision Index (Full) 16.5, median latency 0.04 s. Not in our model list vLLM-SR Decision 2.0 Kai 0.6B: Decision Index (Full) 16.4, median latency 0.02 s. Not in our model list Dinah-1: Decision Index (Full) 16.4, median latency 0.01 s. Not in our model list Kev 0.8B r15: Decision Index (Full) 16.3, median latency 0.04 s. Not in our model list PICK-150M v5: Decision Index (Full) 15.6, median latency 0.02 s. Not in our model list JPT-0.8B: Decision Index (Full) 15.5, median latency 0.12 s. Not in our model list JEV Ettin 400M: Decision Index (Full) 15.2, median latency 0.02 s. Not in our model list Sifr 0.8B v3.1: Decision Index (Full) 14.5, median latency 0.03 s. Not in our model list Crivo-32M v2: Decision Index (Full) 14.2, median latency 0.008 s. Not in our model list Kodiak v0.2 1B (accuracy): Decision Index (Full) 13.9, median latency 0.14 s. Not in our model list Intern-Decision-0.8B: Decision Index (Full) 12.7, median latency 0.03 s. Not in our model list Jev-Style 0.8B v3: Decision Index (Full) 12.5, median latency 0.04 s. Not in our model list Tev1-0.8B-experimental: Decision Index (Full) 12.5, median latency 0.03 s. Not in our model list Bekko System One v0 68M: Decision Index (Full) 12.4, median latency 0.01 s. Not in our model list Amber 17M: Decision Index (Full) 12, median latency 0.006 s. Not in our model list LiquidAI d1-omni-600M: Decision Index (Full) 11.8, median latency 0.01 s. Not in our model list Crivo-17M: Decision Index (Full) 11.7, median latency 0.006 s. Not in our model list Von 1.3: Decision Index (Full) 11.7, median latency 0.02 s. Not in our model list Bosun v3.1 0.6B: Decision Index (Full) 11, median latency 0.04 s. Not in our model list vera-spark: Decision Index (Full) 11, median latency 0.08 s. Not in our model list Lavoir: Decision Index (Full) 10.8, median latency 0.02 s. Not in our model list GLiNER2.5-Decide: Decision Index (Full) 10.1, median latency 0.02 s. Not in our model list Laya: Decision Index (Full) 9.21, median latency 0.006 s. Not in our model list JevK5 Lite: Decision Index (Full) 8.78, median latency 0.02 s. Not in our model list LFM2.5-2.6B-RLCD: Decision Index (Full) 8.18, median latency 0.04 s. Not in our model list Qwen-2.5-1B-RLCD: Decision Index (Full) 7.85, median latency 0.04 s. Not in our model list Bekko System One v0 17M: Decision Index (Full) 7.66, median latency 0.007 s. Not in our model list EmbeddingGemma 2 System-One: Decision Index (Full) 7.41, median latency 0.03 s. Not in our model list CLM-v0.1-8B: Decision Index (Full) 6.49, median latency 0.05 s. Not in our model list jeff: Decision Index (Full) 6.21, median latency 0.02 s. Not in our model list LFM2.5-350M-RLCD: Decision Index (Full) 5.86, median latency 0.03 s. Not in our model list system-one-gemma: Decision Index (Full) 5.22, median latency 0.03 s. Not in our model list DM-JEPA 1.1: Decision Index (Full) 4.96, median latency 0.06 s. Not in our model list Julia 1: Decision Index (Full) 4.75, median latency 0.006 s. Not in our model list Nirnaya 0.4B: Decision Index (Full) 3.51, median latency 0.01 s. Not in our model list Zev: Decision Index (Full) 3.37, median latency 0.0005 s. Not in our model list Verdict: Decision Index (Full) 2.04, median latency 0.01 s. Not in our model list karpo-decision-31B: Decision Index (Full) 73.9, median latency 0.09 s. Not in our model list Cerno 27B v1.1: Decision Index (Full) 69, median latency 0.08 s. Not in our model list AutoTrust-JEV-27B v1.1: Decision Index (Full) 68, median latency 0.08 s. Not in our model list StartLux-Decision v1.1 27B: Decision Index (Full) 67.9, median latency 0.08 s. Not in our model list Decider 31B (NVFP4): Decision Index (Full) 66, median latency 0.07 s. Not in our model list KnowLine 35B-A3B Gen1.5: Decision Index (Full) 65.8, median latency 0.05 s. Not in our model list span-02 (Gemma 4 26B-A4B): Decision Index (Full) 59.1, median latency 0.03 s. Not in our model list Blink v0.3 26B-A4B (NVFP4): Decision Index (Full) 55.9, median latency 0.03 s. Not in our model list H2O Lightning E4B v1.0.1: Decision Index (Full) 48, median latency 0.02 s. Not in our model list Mica v0.5 4B: Decision Index (Full) 47.7, median latency 0.02 s. Not in our model list Decider 4B: Decision Index (Full) 36.3, median latency 0.01 s. Not in our model list Decider 2B: Decision Index (Full) 23.6, median latency 0.008 s. Not in our model list Amber 32M: Decision Index (Full) 16, median latency 0.007 s. Not in our model list Prumo: Decision Index (Full) 14, median latency 0.0003 s. Not in our model list karpo-decision-31BCerno 27B v1.1AutoTrust-JEV-27B v1.1StartLux-Decision v1.1 27BDecider 31B (NVFP4)KnowLine 35B-A3B Gen1.5span-02 (Gemma 4 26B-A4B)Blink v0.3 26B-A4B (NVFP4)H2O Lightning E4B v1.0.1Mica v0.5 4BDecider 4BDecider 2BAmber 32MPrumo 0 20 40 60 80 0.00010.0010.010.11 Median latency (s), log scale, lower is better Decision Index (Full) Gemma 4 31B IT: Decision Index (Full) 61.6, median latency 0.11 s Gemma 4 31B IT: Decision Index (Full) 63.6, median latency 0.1 s Gemma 4 31B IT: Decision Index (Full) 64.6, median latency 0.1 s Gemma 4 31B IT: Decision Index (Full) 67.6, median latency 0.11 s Qwen3.6 27B: Decision Index (Full) 54.7, median latency 0.08 s Qwen 3.8 27B: Decision Index (Full) 57, median latency 0.37 s GLiDE 27B: Decision Index (Full) 73.4, median latency 0.12 s. Not in our model list REAX 1.3: Decision Index (Full) 72.6, median latency 0.25 s. Not in our model list Logits Sigma-1 Duo: Decision Index (Full) 72.3, median latency 0.37 s. Not in our model list Copernicus-31B: Decision Index (Full) 72.1, median latency 0.15 s. Not in our model list KnowLine 31B Gen1.5: Decision Index (Full) 71.9, median latency 0.2 s. Not in our model list Copernicus-57B: Decision Index (Full) 71.8, median latency 0.25 s. Not in our model list FlexDM-LoRA 2.1 · Gemma-4-31B: Decision Index (Full) 71.6, median latency 0.15 s. Not in our model list Sieve 31B v3: Decision Index (Full) 71.1, median latency 0.23 s. Not in our model list OneJev1.6g-31B: Decision Index (Full) 70.9, median latency 0.09 s. Not in our model list Turion Jev 31B: Decision Index (Full) 70.5, median latency 0.14 s. Not in our model list eshbar-lulilla (Gemma 4 31B): Decision Index (Full) 70.3, median latency 0.34 s. Not in our model list Jeffan-27B-1211: Decision Index (Full) 70.3, median latency 0.1 s. Not in our model list vLLM-SR d3-ultra (Decision 3.0, 27B): Decision Index (Full) 70.2, median latency 0.2 s. Not in our model list Bespoke Nimble v4.1 Max (27B): Decision Index (Full) 70, median latency 0.11 s. Not in our model list H2O Lightning 31B v1.3.0: Decision Index (Full) 69.9, median latency 0.14 s. Not in our model list Jeffan-31b-1710: Decision Index (Full) 69.8, median latency 0.12 s. Not in our model list Logits Sigma-1: Decision Index (Full) 69.5, median latency 0.12 s. Not in our model list liquamen (Gemma 4 31B): Decision Index (Full) 69.1, median latency 0.32 s. Not in our model list WeMM-JEV-27B: Decision Index (Full) 68.5, median latency 0.1 s. Not in our model list baselabs mammoth-1 (27B): Decision Index (Full) 68.4, median latency 0.22 s. Not in our model list Piccolo 27B: Decision Index (Full) 68.4, median latency 0.1 s. Not in our model list Ainvest-Decision-27B v1.0: Decision Index (Full) 68.2, median latency 0.1 s. Not in our model list oenogarum (Gemma 4 31B): Decision Index (Full) 68.1, median latency 0.26 s. Not in our model list Perplexity Decider v1.2 (27B): Decision Index (Full) 68, median latency 0.14 s. Not in our model list Perplexity Decider v1.3 (27B): Decision Index (Full) 67.7, median latency 0.14 s. Not in our model list MISA AI 2.0 Decision Plus · oak 1.1 (31B): Decision Index (Full) 67.6, median latency 0.11 s. Not in our model list AJev lm6q (Gemma 4 31B): Decision Index (Full) 67.6, median latency 0.14 s. Not in our model list Chon-1.1-31B: Decision Index (Full) 67.5, median latency 0.17 s. Not in our model list decy-preview (31B): Decision Index (Full) 67.5, median latency 0.12 s. Not in our model list Lychee-27B: Decision Index (Full) 67.3, median latency 0.1 s. Not in our model list vLLM-SR d3-turbo v1.1 (27B): Decision Index (Full) 67.3, median latency 0.19 s. Not in our model list vLLM-SR d3 (27B): Decision Index (Full) 67.2, median latency 0.1 s. Not in our model list allec-oavg (Gemma 4 31B): Decision Index (Full) 67, median latency 0.26 s. Not in our model list VisionJEV 1.0.0 (27B): Decision Index (Full) 66.9, median latency 0.09 s. Not in our model list Aleria Judge 1.1.1 (Gemma 4 31B): Decision Index (Full) 66.7, median latency 0.12 s. Not in our model list Nace.AI Drex Max v1 (27B): Decision Index (Full) 66.6, median latency 0.1 s. Not in our model list AnyJev Tacitus 31B: Decision Index (Full) 66.5, median latency 0.1 s. Not in our model list vera-large-2o: Decision Index (Full) 66.3, median latency 0.15 s. Not in our model list OneShot 27B: Decision Index (Full) 65.7, median latency 0.13 s. Not in our model list Valen-Decision 1.5 27B: Decision Index (Full) 65.2, median latency 0.11 s. Not in our model list Bobcat 1.3 (31B): Decision Index (Full) 64.8, median latency 0.06 s. Not in our model list Amos 1 Large (27B): Decision Index (Full) 64.8, median latency 0.15 s. Not in our model list Matilda Decision v1: Decision Index (Full) 64.8, median latency 0.09 s. Not in our model list KnowLine 26B-A4B Gen2.5: Decision Index (Full) 64.6, median latency 0.07 s. Not in our model list frontier-infra Jebadiah 27B v2.1 (t1e): Decision Index (Full) 64.4, median latency 0.3 s. Not in our model list Darwin-27B-RSI-Decision: Decision Index (Full) 64.2, median latency 0.12 s. Not in our model list deck31b (Gemma 4 31B): Decision Index (Full) 63, median latency 0.22 s. Not in our model list decisio 0.6.0 (Gemma 4 31B): Decision Index (Full) 62.7, median latency 0.05 s. Not in our model list GLiDE 9B: Decision Index (Full) 62.5, median latency 0.09 s. Not in our model list Torchcast Decision 27B: Decision Index (Full) 61.5, median latency 0.1 s. Not in our model list Jeffan-9B-1211: Decision Index (Full) 61.5, median latency 0.08 s. Not in our model list OneJev1.6-27B: Decision Index (Full) 61.5, median latency 0.11 s. Not in our model list Qwen Decision 27B noncausal LoRA: Decision Index (Full) 61.1, median latency 0.13 s. Not in our model list Quyet-1.0-Large: Decision Index (Full) 60.8, median latency 0.11 s. Not in our model list garum (Gemma 4 31B): Decision Index (Full) 60.3, median latency 0.12 s. Not in our model list baselabs elephant-1 (9B): Decision Index (Full) 60.3, median latency 0.11 s. Not in our model list Standard One 35B SH (v1): Decision Index (Full) 60.1, median latency 0.32 s. Not in our model list NTele-Hybrid-27B-v1: Decision Index (Full) 60.1, median latency 0.08 s. Not in our model list Lumma-fev-27B: Decision Index (Full) 60, median latency 0.1 s. Not in our model list Sieve 27B v2: Decision Index (Full) 59.9, median latency 0.11 s. Not in our model list mogan-decision-31B-it: Decision Index (Full) 59.8, median latency 0.11 s. Not in our model list linne-fern 1.0 (27B): Decision Index (Full) 59.6, median latency 0.11 s. Not in our model list Kev 27B: Decision Index (Full) 59.6, median latency 0.14 s. Not in our model list Xor 26B-A4B v1.2: Decision Index (Full) 59.4, median latency 0.11 s. Not in our model list Jev 1.13: Decision Index (Full) 59.2, median latency 0.52 s. Not in our model list Adjudex-9B v5: Decision Index (Full) 59.2, median latency 0.07 s. Not in our model list Cerno 9B v1.2: Decision Index (Full) 59, median latency 0.07 s. Not in our model list garas (Gemma 4 31B): Decision Index (Full) 59, median latency 0.12 s. Not in our model list Shisa DE-2.1 (repeat): Decision Index (Full) 58.8, median latency 0.05 s. Not in our model list Eikos-27B: Decision Index (Full) 58.8, median latency 0.13 s. Not in our model list reflex Qwen3.8-27B-FP8 (pmscr): Decision Index (Full) 58.8, median latency 0.15 s. Not in our model list RSI-Jev v6.1 VL 27B (effort auto): Decision Index (Full) 58.6, median latency 0.08 s. Not in our model list KnowLine 9B Gen1: Decision Index (Full) 58.4, median latency 0.13 s. Not in our model list H2O Lightning Bolt 74B v1.1: Decision Index (Full) 58.4, median latency 0.29 s. Not in our model list StartLux-Decision v1.1 9B: Decision Index (Full) 58.1, median latency 0.04 s. Not in our model list vLLM-SR Decision 2.0 Vega 27B: Decision Index (Full) 58, median latency 0.12 s. Not in our model list JEV-35B: Decision Index (Full) 57.6, median latency 0.05 s. Not in our model list JEV-27B: Decision Index (Full) 57.4, median latency 0.1 s. Not in our model list Piccolo 9B: Decision Index (Full) 57.3, median latency 0.08 s. Not in our model list Tern-35B-A3B: Decision Index (Full) 56.6, median latency 0.06 s. Not in our model list Surogate Rune 26B-A4B v3: Decision Index (Full) 56.4, median latency 0.12 s. Not in our model list SynACK Decide 26B-A4B v7: Decision Index (Full) 56.2, median latency 0.13 s. Not in our model list KnowLine 4B Gen5.5: Decision Index (Full) 56, median latency 0.04 s. Not in our model list Valen-Decision 1.5 9B: Decision Index (Full) 55.9, median latency 0.12 s. Not in our model list JADE: Decision Index (Full) 55.6, median latency 0.13 s. Not in our model list Dexy-Decision-9B-V2: Decision Index (Full) 55.4, median latency 0.03 s. Not in our model list Wald-26B-A4B: Decision Index (Full) 55.4, median latency 0.09 s. Not in our model list AJev lora1 (Gemma 4 26B-A4B): Decision Index (Full) 55.3, median latency 0.09 s. Not in our model list Perplexity Decider Flash v1.3 (4B): Decision Index (Full) 55.1, median latency 0.09 s. Not in our model list Cloudflare clef: Decision Index (Full) 54.9, median latency 0.1 s. Not in our model list StartLux-Decision v1.1 4B: Decision Index (Full) 54.8, median latency 0.04 s. Not in our model list Perplexity Decider Flash v1.2 (4B): Decision Index (Full) 54.3, median latency 0.07 s. Not in our model list vLLM-SR d3-flash (9B): Decision Index (Full) 54.3, median latency 0.05 s. Not in our model list JPT-35B-A3B: Decision Index (Full) 54.1, median latency 0.25 s. Not in our model list chinese-decision-pointer-4b: Decision Index (Full) 54, median latency 0.05 s. Not in our model list Cerno 4B v1.2: Decision Index (Full) 53.9, median latency 0.06 s. Not in our model list Xor 1.2: Decision Index (Full) 53.6, median latency 0.15 s. Not in our model list GEV-26B-Decide: Decision Index (Full) 53.5, median latency 0.13 s. Not in our model list OneJev1.6-9B: Decision Index (Full) 53, median latency 0.04 s. Not in our model list vera-medium-omni: Decision Index (Full) 53, median latency 0.09 s. Not in our model list Senova-4B v4: Decision Index (Full) 52.8, median latency 0.08 s. Not in our model list Nace.AI Drex v2.0 (35B-A3B): Decision Index (Full) 52.8, median latency 0.28 s. Not in our model list muria (Gemma 4 12B): Decision Index (Full) 52.5, median latency 0.05 s. Not in our model list Jeffan-4B-2010: Decision Index (Full) 52.2, median latency 0.05 s. Not in our model list Verdin-4B v1.1: Decision Index (Full) 52, median latency 0.2 s. Not in our model list LatentMesh Blitz 4B v1: Decision Index (Full) 51.9, median latency 0.07 s. Not in our model list Saif 4B v2: Decision Index (Full) 51.7, median latency 0.07 s. Not in our model list vev-4b-v2: Decision Index (Full) 51.1, median latency 0.06 s. Not in our model list Bespoke Nimble 9B v3: Decision Index (Full) 51.1, median latency 0.07 s. Not in our model list frontier-infra Jebadiah 9B v2.1 (t1e): Decision Index (Full) 50.9, median latency 0.13 s. Not in our model list decisio 0.4.0 (Qwen3.6-35B-A3B): Decision Index (Full) 50.8, median latency 0.07 s. Not in our model list Piccolo 4B: Decision Index (Full) 50, median latency 0.05 s. Not in our model list vLLM-SR d3-mini (4B): Decision Index (Full) 50, median latency 0.05 s. Not in our model list RSI-Jev v7.0 VL 4B: Decision Index (Full) 49.9, median latency 0.06 s. Not in our model list Valen-Decision 1.5 4B: Decision Index (Full) 49.9, median latency 0.05 s. Not in our model list Jegue 0.2 (9B): Decision Index (Full) 49.8, median latency 0.09 s. Not in our model list RSI-Jev v7.0 VL 3B: Decision Index (Full) 49.7, median latency 0.1 s. Not in our model list JoshuaSP diffusiongemma (open-jev): Decision Index (Full) 49.5, median latency 0.26 s. Not in our model list AJev lora7 (Gemma 4 12B): Decision Index (Full) 48.9, median latency 0.07 s. Not in our model list OneJev1.6-4B: Decision Index (Full) 48.9, median latency 0.03 s. Not in our model list decisio 0.4.0 (Gemma 4 12B): Decision Index (Full) 48.8, median latency 0.04 s. Not in our model list BaseDecision-Dyno (4B): Decision Index (Full) 48.5, median latency 0.14 s. Not in our model list PurpleMIST-Flash-1.0: Decision Index (Full) 47.5, median latency 0.05 s. Not in our model list Sieve 9B v2: Decision Index (Full) 47, median latency 0.05 s. Not in our model list Decider 35B-A3B: Decision Index (Full) 46.6, median latency 0.1 s. Not in our model list Winnow-12B: Decision Index (Full) 46.6, median latency 0.07 s. Not in our model list Seb-9B v2: Decision Index (Full) 46.3, median latency 0.09 s. Not in our model list PurpleMIST-Mini-2.0: Decision Index (Full) 46.2, median latency 0.06 s. Not in our model list vLLM-SR Decision 2.0 Lux 9B: Decision Index (Full) 46, median latency 0.04 s. Not in our model list Nace.AI Drex v1.5: Decision Index (Full) 45.8, median latency 0.05 s. Not in our model list Hopper 12B (trained): Decision Index (Full) 45.8, median latency 0.29 s. Not in our model list Trio-Spark DI (9B): Decision Index (Full) 45.6, median latency 0.21 s. Not in our model list Jevora 9B v1.6: Decision Index (Full) 45.6, median latency 0.12 s. Not in our model list CoCo-Decision-4B 2.4: Decision Index (Full) 45.5, median latency 0.06 s. Not in our model list Cloudflare clef-omni (30B-A3B): Decision Index (Full) 45, median latency 0.19 s. Not in our model list JevK5 9B v0.3.3: Decision Index (Full) 44.9, median latency 0.04 s. Not in our model list Cloudflare clef-flash: Decision Index (Full) 44.4, median latency 0.05 s. Not in our model list Jevfire: Decision Index (Full) 44.4, median latency 0.09 s. Not in our model list jiwo 4B v2: Decision Index (Full) 44.1, median latency 0.05 s. Not in our model list JPT-9B: Decision Index (Full) 44.1, median latency 0.15 s. Not in our model list ezjev 4B s2: Decision Index (Full) 43.9, median latency 0.04 s. Not in our model list Jev-Omni: Decision Index (Full) 43.8, median latency 0.05 s. Not in our model list GLiDE 2B: Decision Index (Full) 43.7, median latency 0.04 s. Not in our model list Solomon v1.1: Decision Index (Full) 43.2, median latency 0.22 s. Not in our model list Standard One 8B SH (v4): Decision Index (Full) 42.9, median latency 0.09 s. Not in our model list Amber Hybrid 4B: Decision Index (Full) 42.7, median latency 0.27 s. Not in our model list JEV-9B: Decision Index (Full) 42.1, median latency 0.05 s. Not in our model list Bespoke Nimble 9B v2: Decision Index (Full) 42.1, median latency 0.08 s. Not in our model list Sieve 9B: Decision Index (Full) 41.7, median latency 0.05 s. Not in our model list Sieve 4B: Decision Index (Full) 41, median latency 0.09 s. Not in our model list Kev 9B v2: Decision Index (Full) 40.9, median latency 0.06 s. Not in our model list Nari Decision 4B v3.2: Decision Index (Full) 40.9, median latency 0.02 s. Not in our model list Aplomb 1: Decision Index (Full) 40.8, median latency 0.05 s. Not in our model list vLLM-SR Decision 2.0 Nox 4B: Decision Index (Full) 40.8, median latency 0.03 s. Not in our model list Hopper (G) 1.2: Decision Index (Full) 40.2, median latency 0.02 s. Not in our model list StartLux-Decision v1.1 2B: Decision Index (Full) 40.1, median latency 0.03 s. Not in our model list FlyMy.AI Decision 4B v1.6: Decision Index (Full) 39.8, median latency 0.03 s. Not in our model list Standard One 3B SH (v4): Decision Index (Full) 39.8, median latency 0.08 s. Not in our model list JPT-4B: Decision Index (Full) 39.6, median latency 0.14 s. Not in our model list Amber 4B: Decision Index (Full) 39.5, median latency 0.13 s. Not in our model list Intern-Decision-4B: Decision Index (Full) 38.7, median latency 0.04 s. Not in our model list djev: Decision Index (Full) 38.1, median latency 0.08 s. Not in our model list KnowLine 2B Gen1: Decision Index (Full) 38.1, median latency 0.03 s. Not in our model list Valen-Decision 1.5 2B: Decision Index (Full) 38, median latency 0.04 s. Not in our model list razorback16 openjev diffusiongemma (NVFP4, vLLM): Decision Index (Full) 37.6, median latency 0.04 s. Not in our model list NeoHorse-Jev-4B: Decision Index (Full) 37.3, median latency 0.05 s. Not in our model list JevK5 v0.3: Decision Index (Full) 36.6, median latency 0.03 s. Not in our model list GLiDE 0.8B: Decision Index (Full) 36.4, median latency 0.04 s. Not in our model list baselabs joey-1 (2B): Decision Index (Full) 35.9, median latency 0.08 s. Not in our model list lev: Decision Index (Full) 35.5, median latency 0.07 s. Not in our model list Lev1 3.7B: Decision Index (Full) 35.4, median latency 0.03 s. Not in our model list Jev-Style 2B v4.2: Decision Index (Full) 35.1, median latency 0.03 s. Not in our model list Kev 4B r10: Decision Index (Full) 34.9, median latency 0.06 s. Not in our model list PKU-KCL-Meteor-2B: Decision Index (Full) 34.8, median latency 0.03 s. Not in our model list Sieve 2B v3: Decision Index (Full) 34.5, median latency 0.04 s. Not in our model list LatentMesh Blitz 2B v1: Decision Index (Full) 34.4, median latency 0.05 s. Not in our model list Saracura PT-BR 4B: Decision Index (Full) 34.4, median latency 0.06 s. Not in our model list LiquidAI d1-2B: Decision Index (Full) 34.4, median latency 0.09 s. Not in our model list Cam1: Decision Index (Full) 34.4, median latency 0.25 s. Not in our model list vLLM-SR d3-nano (2B): Decision Index (Full) 34.3, median latency 0.04 s. Not in our model list mmastrac diffusiongemma (vLLM PR 57250): Decision Index (Full) 34.3, median latency 0.13 s. Not in our model list Kas 4B: Decision Index (Full) 34, median latency 0.07 s. Not in our model list GLiDE 0.4B: Decision Index (Full) 33.8, median latency 0.02 s. Not in our model list Tev1-4B-experimental: Decision Index (Full) 33.7, median latency 0.04 s. Not in our model list LiquidAI d1-Omni-2B: Decision Index (Full) 33.6, median latency 0.1 s. Not in our model list Adjudex-0.8B v5: Decision Index (Full) 32.8, median latency 0.05 s. Not in our model list Winnow-E4B: Decision Index (Full) 32.7, median latency 0.04 s. Not in our model list Aurelis-2B: Decision Index (Full) 32.5, median latency 0.07 s. Not in our model list GLiClass v4.1 edge (0.4B): Decision Index (Full) 32.5, median latency 0.02 s. Not in our model list Jevstral 8B: Decision Index (Full) 32.2, median latency 0.03 s. Not in our model list Metask-Jev-4B: Decision Index (Full) 32.2, median latency 0.06 s. Not in our model list Ryotide Qwen v0.3.1: Decision Index (Full) 32, median latency 0.08 s. Not in our model list Intelif Qwen3-4B: Decision Index (Full) 31.3, median latency 0.02 s. Not in our model list LiquidAI d1-3B: Decision Index (Full) 31.1, median latency 0.02 s. Not in our model list open-jev (pngwn): Decision Index (Full) 31.1, median latency 0.11 s. Not in our model list PKU-KCL-Meteor-0.8B: Decision Index (Full) 30.9, median latency 0.04 s. Not in our model list vera-core-2: Decision Index (Full) 30.6, median latency 0.09 s. Not in our model list Jobe: Decision Index (Full) 30.4, median latency 0.05 s. Not in our model list OneJev1.6-0.8B: Decision Index (Full) 29.6, median latency 0.01 s. Not in our model list crystal-0.8b: Decision Index (Full) 29.4, median latency 0.05 s. Not in our model list MoDe-L: Decision Index (Full) 29.4, median latency 0.1 s. Not in our model list vLLM-SR vela3-fast (Vela 3.0, 0.4B): Decision Index (Full) 29.3, median latency 0.05 s. Not in our model list Crivo-1B: Decision Index (Full) 29.3, median latency 0.02 s. Not in our model list KnowLine 0.8B Gen1: Decision Index (Full) 29.2, median latency 0.03 s. Not in our model list Gevva E4B: Decision Index (Full) 29.2, median latency 0.05 s. Not in our model list StartLux-Decision v1.1 0.8B: Decision Index (Full) 28.9, median latency 0.03 s. Not in our model list openvons: Decision Index (Full) 28.8, median latency 0.02 s. Not in our model list Amber 0.4B v2: Decision Index (Full) 27.8, median latency 0.02 s. Not in our model list Valen-Decision 1.5 0.8B: Decision Index (Full) 27.6, median latency 0.04 s. Not in our model list SemIf: Decision Index (Full) 26.7, median latency 0.11 s. Not in our model list Saina Helm 2 Omni (0.8B): Decision Index (Full) 26.5, median latency 0.04 s. Not in our model list MoDe-M: Decision Index (Full) 26.5, median latency 0.12 s. Not in our model list Saif 0.8B Lightning v0.3: Decision Index (Full) 26.4, median latency 0.06 s. Not in our model list vLLM-SR d3-lite (0.8B): Decision Index (Full) 26.2, median latency 0.04 s. Not in our model list jiwo 0.8B v2: Decision Index (Full) 25.8, median latency 0.04 s. Not in our model list Saina Helm 2.0 v2: Decision Index (Full) 25.3, median latency 0.04 s. Not in our model list this-that 1.2: Decision Index (Full) 24.9, median latency 0.04 s. Not in our model list Amber 150M: Decision Index (Full) 24.7, median latency 0.02 s. Not in our model list MoDe-S: Decision Index (Full) 24.5, median latency 0.12 s. Not in our model list Gevva E2B: Decision Index (Full) 24.3, median latency 0.04 s. Not in our model list Crivo-400M: Decision Index (Full) 23.7, median latency 0.02 s. Not in our model list vLLM-SR Decision 2.0 Sol 2B: Decision Index (Full) 23.6, median latency 0.03 s. Not in our model list jdi-ptr14-0.8b: Decision Index (Full) 23.6, median latency 0.07 s. Not in our model list mini-jev: Decision Index (Full) 22.8, median latency 0.07 s. Not in our model list vLLM-SR vela3-fast-vision (Vela 3.0, 0.25B): Decision Index (Full) 22.1, median latency 0.04 s. Not in our model list Sieve 2B: Decision Index (Full) 21.9, median latency 0.06 s. Not in our model list wyat-1-800m: Decision Index (Full) 21.5, median latency 0.11 s. Not in our model list Intern-Decision-2B: Decision Index (Full) 20.9, median latency 0.03 s. Not in our model list Sieve 0.8B: Decision Index (Full) 20.8, median latency 0.05 s. Not in our model list Strands Decider 2B: Decision Index (Full) 20.7, median latency 0.05 s. Not in our model list EXAONE-4.0-1.2B-JEV v0.3: Decision Index (Full) 20.6, median latency 0.15 s. Not in our model list tasksource Decider nano v5 (150M): Decision Index (Full) 19.9, median latency 0.01 s. Not in our model list Crivo-68M: Decision Index (Full) 19.5, median latency 0.01 s. Not in our model list Amber 68M: Decision Index (Full) 19.2, median latency 0.01 s. Not in our model list JEV Ettin 150M: Decision Index (Full) 18.5, median latency 0.08 s. Not in our model list vLLM-SR Decision 2.0 Eos 0.8B: Decision Index (Full) 18.4, median latency 0.03 s. Not in our model list Crivo-150M v2: Decision Index (Full) 18.3, median latency 0.01 s. Not in our model list Bekko System One v0 400M: Decision Index (Full) 18, median latency 0.02 s. Not in our model list Mola-150M: Decision Index (Full) 17.8, median latency 0.01 s. Not in our model list vLLM-SR d3-edge (0.6B): Decision Index (Full) 17.7, median latency 0.02 s. Not in our model list Mola-68M: Decision Index (Full) 17.6, median latency 0.01 s. Not in our model list JEV FLAN-T5 Large public v2 (342M): Decision Index (Full) 17.3, median latency 0.02 s. Not in our model list DecisionTune 1.0: Decision Index (Full) 17.2, median latency 0.01 s. Not in our model list Bosun v3.1 1.7B: Decision Index (Full) 16.5, median latency 0.04 s. Not in our model list vLLM-SR Decision 2.0 Kai 0.6B: Decision Index (Full) 16.4, median latency 0.02 s. Not in our model list Dinah-1: Decision Index (Full) 16.4, median latency 0.01 s. Not in our model list Kev 0.8B r15: Decision Index (Full) 16.3, median latency 0.04 s. Not in our model list PICK-150M v5: Decision Index (Full) 15.6, median latency 0.02 s. Not in our model list JPT-0.8B: Decision Index (Full) 15.5, median latency 0.12 s. Not in our model list JEV Ettin 400M: Decision Index (Full) 15.2, median latency 0.02 s. Not in our model list Sifr 0.8B v3.1: Decision Index (Full) 14.5, median latency 0.03 s. Not in our model list Crivo-32M v2: Decision Index (Full) 14.2, median latency 0.008 s. Not in our model list Kodiak v0.2 1B (accuracy): Decision Index (Full) 13.9, median latency 0.14 s. Not in our model list Intern-Decision-0.8B: Decision Index (Full) 12.7, median latency 0.03 s. Not in our model list Jev-Style 0.8B v3: Decision Index (Full) 12.5, median latency 0.04 s. Not in our model list Tev1-0.8B-experimental: Decision Index (Full) 12.5, median latency 0.03 s. Not in our model list Bekko System One v0 68M: Decision Index (Full) 12.4, median latency 0.01 s. Not in our model list Amber 17M: Decision Index (Full) 12, median latency 0.006 s. Not in our model list LiquidAI d1-omni-600M: Decision Index (Full) 11.8, median latency 0.01 s. Not in our model list Crivo-17M: Decision Index (Full) 11.7, median latency 0.006 s. Not in our model list Von 1.3: Decision Index (Full) 11.7, median latency 0.02 s. Not in our model list Bosun v3.1 0.6B: Decision Index (Full) 11, median latency 0.04 s. Not in our model list vera-spark: Decision Index (Full) 11, median latency 0.08 s. Not in our model list Lavoir: Decision Index (Full) 10.8, median latency 0.02 s. Not in our model list GLiNER2.5-Decide: Decision Index (Full) 10.1, median latency 0.02 s. Not in our model list Laya: Decision Index (Full) 9.21, median latency 0.006 s. Not in our model list JevK5 Lite: Decision Index (Full) 8.78, median latency 0.02 s. Not in our model list LFM2.5-2.6B-RLCD: Decision Index (Full) 8.18, median latency 0.04 s. Not in our model list Qwen-2.5-1B-RLCD: Decision Index (Full) 7.85, median latency 0.04 s. Not in our model list Bekko System One v0 17M: Decision Index (Full) 7.66, median latency 0.007 s. Not in our model list EmbeddingGemma 2 System-One: Decision Index (Full) 7.41, median latency 0.03 s. Not in our model list CLM-v0.1-8B: Decision Index (Full) 6.49, median latency 0.05 s. Not in our model list jeff: Decision Index (Full) 6.21, median latency 0.02 s. Not in our model list LFM2.5-350M-RLCD: Decision Index (Full) 5.86, median latency 0.03 s. Not in our model list system-one-gemma: Decision Index (Full) 5.22, median latency 0.03 s. Not in our model list DM-JEPA 1.1: Decision Index (Full) 4.96, median latency 0.06 s. Not in our model list Julia 1: Decision Index (Full) 4.75, median latency 0.006 s. Not in our model list Nirnaya 0.4B: Decision Index (Full) 3.51, median latency 0.01 s. Not in our model list Zev: Decision Index (Full) 3.37, median latency 0.0005 s. Not in our model list Verdict: Decision Index (Full) 2.04, median latency 0.01 s. Not in our model list karpo-decision-31B: Decision Index (Full) 73.9, median latency 0.09 s. Not in our model list Cerno 27B v1.1: Decision Index (Full) 69, median latency 0.08 s. Not in our model list AutoTrust-JEV-27B v1.1: Decision Index (Full) 68, median latency 0.08 s. Not in our model list StartLux-Decision v1.1 27B: Decision Index (Full) 67.9, median latency 0.08 s. Not in our model list Decider 31B (NVFP4): Decision Index (Full) 66, median latency 0.07 s. Not in our model list KnowLine 35B-A3B Gen1.5: Decision Index (Full) 65.8, median latency 0.05 s. Not in our model list span-02 (Gemma 4 26B-A4B): Decision Index (Full) 59.1, median latency 0.03 s. Not in our model list Blink v0.3 26B-A4B (NVFP4): Decision Index (Full) 55.9, median latency 0.03 s. Not in our model list H2O Lightning E4B v1.0.1: Decision Index (Full) 48, median latency 0.02 s. Not in our model list Mica v0.5 4B: Decision Index (Full) 47.7, median latency 0.02 s. Not in our model list Decider 4B: Decision Index (Full) 36.3, median latency 0.01 s. Not in our model list Decider 2B: Decision Index (Full) 23.6, median latency 0.008 s. Not in our model list Amber 32M: Decision Index (Full) 16, median latency 0.007 s. Not in our model list Prumo: Decision Index (Full) 14, median latency 0.0003 s. Not in our model list karpo-decision-31BCerno 27B v1.1AutoTrust-JEV-27B v1.1StartLux-Decision v1.1 27BDecider 31B (NVFP4)KnowLine 35B-A3B Gen1.5span-02 (Gemma 4 26B-A4B)Blink v0.3 26B-A4B (NVFP4)H2O Lightning E4B v1.0.1Mica v0.5 4BDecider 4BDecider 2BAmber 32MPrumo
ModelSettingDecision Index (Full)Median latency
karpo-decision-31B 73.86 0.09 s
GLiDE 27B 73.37 0.12 s
REAX 1.3 72.64 0.25 s
Logits Sigma-1 Duo 72.28 0.37 s
Copernicus-31B 72.11 0.15 s
KnowLine 31B Gen1.5 71.91 0.2 s
Copernicus-57B 71.8 0.25 s
FlexDM-LoRA 2.1 · Gemma-4-31B 71.61 0.16 s
Sieve 31B v3 71.1 0.23 s
OneJev1.6g-31B 70.86 0.09 s
Turion Jev 31B 70.47 0.14 s
Jeffan-27B-1211 70.3 0.1 s
eshbar-lulilla (Gemma 4 31B) 70.3 0.34 s
vLLM-SR d3-ultra (Decision 3.0, 27B) 70.18 0.2 s
Bespoke Nimble v4.1 Max (27B) 70.03 0.11 s
H2O Lightning 31B v1.3.0 69.92 0.14 s
Jeffan-31b-1710 69.75 0.12 s
Logits Sigma-1 69.47 0.12 s
liquamen (Gemma 4 31B) 69.14 0.32 s
Cerno 27B v1.1 68.96 0.08 s
WeMM-JEV-27B 68.49 0.1 s
baselabs mammoth-1 (27B) 68.44 0.22 s
Piccolo 27B 68.4 0.1 s
Ainvest-Decision-27B v1.0 68.23 0.1 s
oenogarum (Gemma 4 31B) 68.1 0.26 s
Perplexity Decider v1.2 (27B) 67.98 0.14 s
AutoTrust-JEV-27B v1.1 67.95 0.08 s
StartLux-Decision v1.1 27B 67.92 0.08 s
Perplexity Decider v1.3 (27B) 67.67 0.14 s
MISA AI 2.0 Decision Plus · oak 1.1 (31B) 67.62 0.11 s
AJev lm6q (Gemma 4 31B) 67.57 0.14 s
Gemma 4 31B IT FlexDM 67.56 0.11 s
Chon-1.1-31B 67.54 0.17 s
decy-preview (31B) 67.48 0.12 s
Lychee-27B 67.35 0.1 s
vLLM-SR d3-turbo v1.1 (27B) 67.31 0.19 s
vLLM-SR d3 (27B) 67.15 0.1 s
allec-oavg (Gemma 4 31B) 66.96 0.26 s
VisionJEV 1.0.0 (27B) 66.88 0.09 s
Aleria Judge 1.1.1 (Gemma 4 31B) 66.67 0.12 s
Nace.AI Drex Max v1 (27B) 66.6 0.1 s
AnyJev Tacitus 31B 66.47 0.1 s
vera-large-2o 66.26 0.15 s
Decider 31B (NVFP4) 66.03 0.07 s
KnowLine 35B-A3B Gen1.5 65.77 0.05 s
OneShot 27B 65.74 0.13 s
Valen-Decision 1.5 27B 65.21 0.11 s
Bobcat 1.3 (31B) 64.85 0.06 s
Amos 1 Large (27B) 64.81 0.15 s
Matilda Decision v1 64.75 0.09 s
Gemma 4 31B IT jev-brain 64.61 0.1 s
KnowLine 26B-A4B Gen2.5 64.57 0.07 s
frontier-infra Jebadiah 27B v2.1 (t1e) 64.4 0.3 s
Darwin-27B-RSI-Decision 64.18 0.12 s
Gemma 4 31B IT AnyJev L0 63.65 0.1 s
deck31b (Gemma 4 31B) 63.05 0.22 s
decisio 0.6.0 (Gemma 4 31B) 62.68 0.05 s
GLiDE 9B 62.48 0.09 s
Gemma 4 31B IT Decider chat 61.63 0.11 s
Torchcast Decision 27B 61.49 0.1 s
Jeffan-9B-1211 61.48 0.08 s
OneJev1.6-27B 61.46 0.11 s
Qwen Decision 27B noncausal LoRA 61.12 0.13 s
Quyet-1.0-Large 60.77 0.11 s
garum (Gemma 4 31B) 60.27 0.12 s
baselabs elephant-1 (9B) 60.26 0.11 s
Standard One 35B SH (v1) 60.14 0.32 s
NTele-Hybrid-27B-v1 60.12 0.08 s
Lumma-fev-27B 59.97 0.1 s
Sieve 27B v2 59.89 0.11 s
mogan-decision-31B-it 59.82 0.11 s
linne-fern 1.0 (27B) 59.65 0.11 s
Kev 27B 59.56 0.14 s
Xor 26B-A4B v1.2 59.44 0.11 s
Jev 1.13 59.22 0.52 s
Adjudex-9B v5 59.19 0.07 s
span-02 (Gemma 4 26B-A4B) 59.07 0.03 s
Cerno 9B v1.2 58.98 0.07 s
garas (Gemma 4 31B) 58.95 0.12 s
Shisa DE-2.1 (repeat) 58.84 0.05 s
Eikos-27B 58.79 0.13 s
reflex Qwen3.8-27B-FP8 (pmscr) 58.77 0.15 s
RSI-Jev v6.1 VL 27B (effort auto) 58.6 0.08 s
KnowLine 9B Gen1 58.44 0.13 s
H2O Lightning Bolt 74B v1.1 58.42 0.29 s
StartLux-Decision v1.1 9B 58.07 0.04 s
vLLM-SR Decision 2.0 Vega 27B 58.01 0.12 s
JEV-35B 57.61 0.05 s
JEV-27B 57.35 0.1 s
Piccolo 9B 57.34 0.08 s
Qwen 3.8 27B simple-jev 56.99 0.37 s
Tern-35B-A3B 56.62 0.06 s
Surogate Rune 26B-A4B v3 56.42 0.12 s
SynACK Decide 26B-A4B v7 56.22 0.13 s
KnowLine 4B Gen5.5 55.99 0.04 s
Blink v0.3 26B-A4B (NVFP4) 55.92 0.03 s
Valen-Decision 1.5 9B 55.85 0.12 s
JADE 55.57 0.13 s
Dexy-Decision-9B-V2 55.39 0.03 s
Wald-26B-A4B 55.35 0.09 s
AJev lora1 (Gemma 4 26B-A4B) 55.29 0.09 s
Perplexity Decider Flash v1.3 (4B) 55.14 0.09 s
Cloudflare clef 54.92 0.1 s
StartLux-Decision v1.1 4B 54.8 0.04 s
Qwen3.6 27B Decider chat 54.67 0.08 s
Perplexity Decider Flash v1.2 (4B) 54.29 0.07 s
vLLM-SR d3-flash (9B) 54.27 0.05 s
JPT-35B-A3B 54.09 0.25 s
chinese-decision-pointer-4b 53.96 0.05 s
Cerno 4B v1.2 53.88 0.06 s
Xor 1.2 53.62 0.15 s
GEV-26B-Decide 53.5 0.13 s
OneJev1.6-9B 53.05 0.04 s
vera-medium-omni 53.02 0.09 s
Senova-4B v4 52.79 0.08 s
Nace.AI Drex v2.0 (35B-A3B) 52.76 0.28 s
muria (Gemma 4 12B) 52.55 0.05 s
Jeffan-4B-2010 52.21 0.05 s
Verdin-4B v1.1 51.99 0.2 s
LatentMesh Blitz 4B v1 51.94 0.07 s
Saif 4B v2 51.68 0.07 s
vev-4b-v2 51.11 0.06 s
Bespoke Nimble 9B v3 51.06 0.07 s
frontier-infra Jebadiah 9B v2.1 (t1e) 50.9 0.13 s
decisio 0.4.0 (Qwen3.6-35B-A3B) 50.82 0.07 s
Piccolo 4B 50 0.05 s
vLLM-SR d3-mini (4B) 49.97 0.05 s
RSI-Jev v7.0 VL 4B 49.93 0.06 s
Valen-Decision 1.5 4B 49.87 0.05 s
Jegue 0.2 (9B) 49.82 0.09 s
RSI-Jev v7.0 VL 3B 49.68 0.1 s
JoshuaSP diffusiongemma (open-jev) 49.53 0.26 s
AJev lora7 (Gemma 4 12B) 48.91 0.07 s
OneJev1.6-4B 48.88 0.03 s
decisio 0.4.0 (Gemma 4 12B) 48.75 0.04 s
BaseDecision-Dyno (4B) 48.52 0.14 s
H2O Lightning E4B v1.0.1 47.99 0.02 s
Mica v0.5 4B 47.7 0.02 s
PurpleMIST-Flash-1.0 47.54 0.05 s
Sieve 9B v2 47 0.05 s
Decider 35B-A3B 46.64 0.1 s
Winnow-12B 46.57 0.07 s
Seb-9B v2 46.25 0.09 s
PurpleMIST-Mini-2.0 46.24 0.06 s
vLLM-SR Decision 2.0 Lux 9B 45.97 0.04 s
Nace.AI Drex v1.5 45.78 0.05 s
Hopper 12B (trained) 45.76 0.29 s
Trio-Spark DI (9B) 45.59 0.21 s
Jevora 9B v1.6 45.57 0.12 s
CoCo-Decision-4B 2.4 45.52 0.06 s
Cloudflare clef-omni (30B-A3B) 45.04 0.19 s
JevK5 9B v0.3.3 44.89 0.04 s
Cloudflare clef-flash 44.43 0.05 s
Jevfire 44.39 0.09 s
jiwo 4B v2 44.1 0.05 s
JPT-9B 44.08 0.15 s
ezjev 4B s2 43.9 0.04 s
Jev-Omni 43.83 0.05 s
GLiDE 2B 43.72 0.04 s
Solomon v1.1 43.23 0.22 s
Standard One 8B SH (v4) 42.93 0.09 s
Amber Hybrid 4B 42.74 0.27 s
JEV-9B 42.15 0.05 s
Bespoke Nimble 9B v2 42.06 0.08 s
Sieve 9B 41.73 0.05 s
Sieve 4B 40.96 0.09 s
Kev 9B v2 40.91 0.06 s
Nari Decision 4B v3.2 40.87 0.02 s
vLLM-SR Decision 2.0 Nox 4B 40.75 0.03 s
Aplomb 1 40.75 0.05 s
Hopper (G) 1.2 40.17 0.02 s
StartLux-Decision v1.1 2B 40.11 0.03 s
FlyMy.AI Decision 4B v1.6 39.8 0.03 s
Standard One 3B SH (v4) 39.79 0.08 s
JPT-4B 39.56 0.14 s
Amber 4B 39.45 0.13 s
Intern-Decision-4B 38.69 0.04 s
djev 38.15 0.08 s
KnowLine 2B Gen1 38.07 0.03 s
Valen-Decision 1.5 2B 37.97 0.04 s
razorback16 openjev diffusiongemma (NVFP4, vLLM) 37.6 0.04 s
NeoHorse-Jev-4B 37.26 0.05 s
JevK5 v0.3 36.59 0.03 s
GLiDE 0.8B 36.36 0.04 s
Decider 4B 36.3 0.01 s
baselabs joey-1 (2B) 35.91 0.08 s
lev 35.47 0.07 s
Lev1 3.7B 35.41 0.03 s
Jev-Style 2B v4.2 35.13 0.03 s
Kev 4B r10 34.93 0.06 s
PKU-KCL-Meteor-2B 34.75 0.03 s
Sieve 2B v3 34.45 0.04 s
LatentMesh Blitz 2B v1 34.44 0.05 s
Saracura PT-BR 4B 34.43 0.06 s
LiquidAI d1-2B 34.39 0.09 s
Cam1 34.36 0.25 s
vLLM-SR d3-nano (2B) 34.3 0.04 s
mmastrac diffusiongemma (vLLM PR 57250) 34.25 0.13 s
Kas 4B 34 0.07 s
GLiDE 0.4B 33.84 0.02 s
Tev1-4B-experimental 33.73 0.04 s
LiquidAI d1-Omni-2B 33.58 0.1 s
Adjudex-0.8B v5 32.81 0.05 s
Winnow-E4B 32.72 0.05 s
Aurelis-2B 32.55 0.07 s
GLiClass v4.1 edge (0.4B) 32.47 0.02 s
Jevstral 8B 32.23 0.03 s
Metask-Jev-4B 32.17 0.06 s
Ryotide Qwen v0.3.1 31.99 0.08 s
Intelif Qwen3-4B 31.33 0.02 s
LiquidAI d1-3B 31.12 0.02 s
open-jev (pngwn) 31.1 0.11 s
PKU-KCL-Meteor-0.8B 30.94 0.04 s
vera-core-2 30.6 0.09 s
Jobe 30.35 0.05 s
OneJev1.6-0.8B 29.58 0.01 s
crystal-0.8b 29.37 0.05 s
MoDe-L 29.35 0.1 s
vLLM-SR vela3-fast (Vela 3.0, 0.4B) 29.27 0.05 s
Crivo-1B 29.25 0.02 s
KnowLine 0.8B Gen1 29.24 0.03 s
Gevva E4B 29.21 0.05 s
StartLux-Decision v1.1 0.8B 28.88 0.03 s
openvons 28.78 0.02 s
Amber 0.4B v2 27.77 0.02 s
Valen-Decision 1.5 0.8B 27.6 0.04 s
SemIf 26.71 0.11 s
Saina Helm 2 Omni (0.8B) 26.54 0.04 s
MoDe-M 26.49 0.12 s
Saif 0.8B Lightning v0.3 26.37 0.06 s
vLLM-SR d3-lite (0.8B) 26.16 0.04 s
jiwo 0.8B v2 25.81 0.04 s
Saina Helm 2.0 v2 25.29 0.04 s
this-that 1.2 24.85 0.04 s
Amber 150M 24.67 0.02 s
MoDe-S 24.52 0.12 s
Gevva E2B 24.26 0.04 s
Crivo-400M 23.74 0.02 s
Decider 2B 23.65 0.01 s
vLLM-SR Decision 2.0 Sol 2B 23.64 0.03 s
jdi-ptr14-0.8b 23.58 0.07 s
mini-jev 22.81 0.07 s
vLLM-SR vela3-fast-vision (Vela 3.0, 0.25B) 22.11 0.04 s
Sieve 2B 21.9 0.06 s
wyat-1-800m 21.49 0.11 s
Intern-Decision-2B 20.93 0.03 s
Sieve 0.8B 20.82 0.05 s
Strands Decider 2B 20.72 0.05 s
EXAONE-4.0-1.2B-JEV v0.3 20.6 0.15 s
tasksource Decider nano v5 (150M) 19.85 0.01 s
Crivo-68M 19.51 0.01 s
Amber 68M 19.24 0.01 s
JEV Ettin 150M 18.52 0.08 s
vLLM-SR Decision 2.0 Eos 0.8B 18.37 0.03 s
Crivo-150M v2 18.26 0.01 s
Bekko System One v0 400M 17.98 0.02 s
Mola-150M 17.84 0.01 s
vLLM-SR d3-edge (0.6B) 17.69 0.02 s
Mola-68M 17.6 0.01 s
JEV FLAN-T5 Large public v2 (342M) 17.31 0.02 s
DecisionTune 1.0 17.22 0.01 s
Bosun v3.1 1.7B 16.54 0.04 s
vLLM-SR Decision 2.0 Kai 0.6B 16.43 0.02 s
Dinah-1 16.36 0.01 s
Kev 0.8B r15 16.3 0.05 s
Amber 32M 15.98 0.01 s
PICK-150M v5 15.62 0.02 s
JPT-0.8B 15.46 0.12 s
JEV Ettin 400M 15.24 0.02 s
Sifr 0.8B v3.1 14.53 0.03 s
Crivo-32M v2 14.23 0.01 s
Prumo 13.99 0 s
Kodiak v0.2 1B (accuracy) 13.86 0.14 s
Intern-Decision-0.8B 12.71 0.03 s
Jev-Style 0.8B v3 12.5 0.04 s
Tev1-0.8B-experimental 12.47 0.03 s
Bekko System One v0 68M 12.4 0.01 s
Amber 17M 11.99 0.01 s
LiquidAI d1-omni-600M 11.75 0.01 s
Crivo-17M 11.65 0.01 s
Von 1.3 11.65 0.02 s
Bosun v3.1 0.6B 11.04 0.04 s
vera-spark 11 0.08 s
Lavoir 10.85 0.02 s
GLiNER2.5-Decide 10.14 0.02 s
Laya 9.21 0.01 s
JevK5 Lite 8.78 0.02 s
LFM2.5-2.6B-RLCD 8.18 0.04 s
Qwen-2.5-1B-RLCD 7.85 0.04 s
Bekko System One v0 17M 7.66 0.01 s
EmbeddingGemma 2 System-One 7.41 0.03 s
CLM-v0.1-8B 6.49 0.05 s
jeff 6.21 0.02 s
LFM2.5-350M-RLCD 5.86 0.03 s
system-one-gemma 5.22 0.03 s
DM-JEPA 1.1 4.96 0.06 s
Julia 1 4.75 0.01 s
Nirnaya 0.4B 3.51 0.01 s
Zev 3.37 0 s
Verdict 2.04 0.01 s

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind
Own leaderboard Not tested since Sep 2026

Source: Official Decision Index leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
50 55 60 65 70 2026 50 55 60 65 70 2026

Source: Official Decision Index leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Other

All 43 in Other
  • Personal Superintelligence Index (Meta) In use Meta's composite of seven personal and knowledge-work evals: Agents' Last Exam, AutomationBench-AA, FocusBench, JobBench, GDP.pdf, VBT and DRACO.
  • Artificial Analysis Intelligence Index v5 In use Version 5 of Artificial Analysis's index: nine evaluations in four equal categories, most with private test sets, run three times each in its own agent harness. Gemini 4 Carbon 59.0
  • SAGE v1.1 (Vals) In use Models grade handwritten undergraduate math and computer science work against rubrics. Scored by class-balanced accuracy of each rubric check against expert grading. Claude Opus 5.5 91.3%
  • MTEB Fading Massive Text Embedding Benchmark: retrieval, clustering, classification across languages and modalities. sonar 42%
  • Anthropic ECI (AECI) Only in system cards Index that combines a basket of capability benchmarks into one capability score to track trends across Anthropic models. Claude Opus 5.5 169.4
  • SPEED-Bench Only in system cards Measures speculative decoding across varied prompts and domains. Reports the average acceptance length of draft tokens, so higher means faster generation at the same output. Nemotron 3 Ultra 5.15