Benchmark Almanac

CyberBench v1.1 (Vals)

Muse Wonder 80.4% leads since October 2026
In use

Agents work on OSS-Fuzz memory-safety bugs from ARVO images in two tracks. They write a proof-of-concept input that crashes the target, or patch the source so the crash stops and tests still pass.

Vals reran the benchmark with offline task sandboxes, a hardened clean-room grader and a new Patch roster, and archived v1. Release month is inferred: v1 last updated September 16th, 2026, v1.1 shows October 20th, 2026.

Made by
Vals AI
Released
October 2026
Kind
Finding and exploiting real bugs
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 12th, 2026
47
Entries on its own leaderboard, 44 models we track
1
Launch post lists it, from 1 lab

Muse Wonder leads with 80.4%

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. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

47 entries. Scroll sideways for the rest.

  1. 80.4 Muse Wonder
  2. 78.0 GPT-6 Sol
  3. 77.9 Gemini 4 Argon
  4. 76.3 GPT-5.6 Sol
  5. 76.3 GPT-6 Luna
  6. 75.5 Claude Haiku 5.5
  7. 75.4 MiMo-V2.6 Flash
  8. 75.2 Kimi K3
  9. 74.6 Claude Opus 5.5 (CVP)
  10. 73.7 DeepSeek V4.1 Flash
  11. 73.6 GPT-5.6 Luna
  12. 73.6 Muse Wonder Full
  13. 73.0 Daybreak Blue
  14. 72.9 DeepSeek V4 Pro 0813
  15. 72.9 MiMo-V2.6 Pro
  16. 72.7 Muse Spark 1.3
  17. 72.7 Inkling 2 Small
  18. 72.1 GPT-5.6 Terra
  19. 72.1 DeepSeek V4 Flash 0731
  20. 72.1 GLM-5.3
  21. 71.0 Grok 4.5
  22. 70.4 Claude Fable 5.1
  23. 70.3 Claude Sonnet 5 (Cyber Range)
  24. 69.5 Muse Spark 1.2
  25. 69.5 Ember-1
  26. 69.5 Grok 4.7
  27. 67.9 GLM-5.3-Flash
  28. 67.8 Qwen 3.8 27B
  29. 67.7 Grok 4.6
  30. 66.8 Inkling
  31. 65.4 Claude Opus 5
  32. 65.4 Tencent Hy4 preview
  33. 65.0 MiniMax-M3
  34. 64.1 Inkling-Small
  35. 61.9 Claude Sonnet 5
  36. 59.6 Claude Sonnet 5.5
  37. 59.1 Gemini 3.5 Flash-Lite
  38. 58.3 Ling-3.0-flash-Fin
  39. 55.4 Claude Opus 5.5
  40. 45.5 Mercury 2.5
  41. 45.4 Gemini 3.6 Flash
  42. 43.8 Gemini 3.7 Flash
  43. 43.8 Gemini 3.8 Flash
  44. 41.1 GPT-6 Astra
  45. 39.3 GPT-6.1 Sol
  46. 29.5 Nemotron 3.5 Lightning 30B-A3B
  47. 28.6 Qwen 3.8 Max
  1. 1 Muse Wonder 80.4%
  2. 2 GPT-6 Sol 78%
  3. 3 Gemini 4 Argon 77.9%
  4. 4 GPT-5.6 Sol 76.3%
  5. 5 GPT-6 Luna 76.3%
  6. 6 Claude Haiku 5.5 75.5%
  7. 7 MiMo-V2.6 Flash 75.4%
  8. 8 Kimi K3 75.2%
  9. 9 Claude Opus 5.5 (CVP) 74.6%
  10. 10 DeepSeek V4.1 Flash 73.7%
  11. 11 GPT-5.6 Luna 73.6%
  12. 12 Muse Wonder Full 73.6%
  13. 13 Daybreak Blue 73%
  14. 14 DeepSeek V4 Pro 0813 72.9%
  15. 15 MiMo-V2.6 Pro 72.9%
  16. 16 Muse Spark 1.3 max 72.7%
  17. 17 Inkling 2 Small 72.7%
  18. 18 GPT-5.6 Terra 72.1%
  19. 19 DeepSeek V4 Flash 0731 72.1%
  20. 20 GLM-5.3 72.1%
  21. 21 Grok 4.5 71%
  22. 22 Claude Fable 5.1 70.4%
  23. 23 Claude Sonnet 5 (Cyber Range) 70.3%
  24. 24 Muse Spark 1.2 69.5%
  25. 25 Ember-1 69.5%
  26. 26 Grok 4.7 69.5%
  27. 27 GLM-5.3-Flash 67.9%
  28. 28 Qwen 3.8 27B 67.8%
  29. 29 Grok 4.6 67.7%
  30. 30 Inkling 66.8%
  31. 31 Claude Opus 5 65.4%
  32. 32 Tencent Hy4 preview 65.4%
  33. 33 MiniMax-M3 65%
  34. 34 Inkling-Small 64.1%
  35. 35 Claude Sonnet 5 61.9%
  36. 36 Claude Sonnet 5.5 59.6%
  37. 37 Gemini 3.5 Flash-Lite 59.1%
  38. 38 Ling-3.0-flash-Fin 58.3%
  39. 39 Claude Opus 5.5 55.4%
  40. 40 Mercury 2.5 45.5%
  41. 41 Gemini 3.6 Flash 45.4%
  42. 42 Gemini 3.7 Flash 43.8%
  43. 43 Gemini 3.8 Flash 43.8%
  44. 44 GPT-6 Astra 41.1%
  45. 45 GPT-6.1 Sol 39.3%
  46. 46 Nemotron 3.5 Lightning 30B-A3B 29.5%
  47. 47 Qwen 3.8 Max 28.6%

Source: Official CyberBench v1.1 (Vals) leaderboard

#EntryScoreSettingDateSource
1 Muse Wonder 80.4% Oct 2026 Official leaderboard
2 GPT-6 Sol 78% Sep 2026 Official leaderboard
3 Gemini 4 Argon 77.9% Sep 2026 Official leaderboard
4 GPT-5.6 Sol 76.3% Jul 2026 Official leaderboard
5 GPT-6 Luna 76.3% Sep 2026 Official leaderboard
6 Claude Haiku 5.5 75.5% Oct 2026 Official leaderboard
7 MiMo-V2.6 Flash 75.4% Sep 2026 Official leaderboard
8 Kimi K3 75.2% Jul 2026 Official leaderboard
9 Claude Opus 5.5 (CVP) 74.6% Official leaderboard
10 DeepSeek V4.1 Flash 73.7% Sep 2026 Official leaderboard
11 GPT-5.6 Luna 73.6% Jul 2026 Official leaderboard
12 Muse Wonder Full 73.6% Oct 2026 Official leaderboard
13 Daybreak Blue 73% Official leaderboard
14 DeepSeek V4 Pro 0813 72.9% Aug 2026 Official leaderboard
15 MiMo-V2.6 Pro 72.9% Sep 2026 Official leaderboard
16 Muse Spark 1.3 72.7% max Sep 2026 Official leaderboard
17 Inkling 2 Small 72.7% Nov 2026 Official leaderboard
18 GPT-5.6 Terra 72.1% Jul 2026 Official leaderboard
19 DeepSeek V4 Flash 0731 72.1% Jul 2026 Official leaderboard
20 GLM-5.3 72.1% Aug 2026 Official leaderboard
21 Grok 4.5 71% Jul 2026 Official leaderboard
22 Claude Fable 5.1 70.4% Sep 2026 Official leaderboard
23 Claude Sonnet 5 (Cyber Range) 70.3% Official leaderboard
24 Muse Spark 1.2 69.5% Aug 2026 Official leaderboard
25 Ember-1 69.5% Sep 2026 Official leaderboard
26 Grok 4.7 69.5% Sep 2026 Official leaderboard
27 GLM-5.3-Flash 67.9% Aug 2026 Official leaderboard
28 Qwen 3.8 27B 67.8% Aug 2026 Official leaderboard
29 Grok 4.6 67.7% Aug 2026 Official leaderboard
30 Inkling 66.8% Jul 2026 Official leaderboard
31 Claude Opus 5 65.4% Jul 2026 Official leaderboard
32 Tencent Hy4 preview 65.4% Aug 2026 Official leaderboard
33 MiniMax-M3 65% Jun 2026 Official leaderboard
34 Inkling-Small 64.1% Jul 2026 Official leaderboard
35 Claude Sonnet 5 61.9% Jun 2026 Official leaderboard
36 Claude Sonnet 5.5 59.6% Sep 2026 Official leaderboard
37 Gemini 3.5 Flash-Lite 59.1% Jul 2026 Official leaderboard
38 Ling-3.0-flash-Fin 58.3% Sep 2026 Official leaderboard
39 Claude Opus 5.5 55.4% Sep 2026 Official leaderboard
40 Mercury 2.5 45.5% Sep 2026 Official leaderboard
41 Gemini 3.6 Flash 45.4% Jul 2026 Official leaderboard
42 Gemini 3.7 Flash 43.8% Aug 2026 Official leaderboard
43 Gemini 3.8 Flash 43.8% Sep 2026 Official leaderboard
44 GPT-6 Astra 41.1% Sep 2026 Official leaderboard
45 GPT-6.1 Sol 39.3% Sep 2026 Official leaderboard
46 Nemotron 3.5 Lightning 30B-A3B 29.5% Aug 2026 Official leaderboard
47 Qwen 3.8 Max 28.6% Aug 2026 Official leaderboard

Labs report scores for 1 model in their cards. We plot every source in the charts below.

Accuracy against latency

The Vals scatter chart with its x axis switched from Cost to Latency: mean time per test against accuracy, on a log scale. 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% 100% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy MiniMax-M3: Accuracy 65%, latency (mean per test) 2,456 s Claude Sonnet 5: Accuracy 61.9%, latency (mean per test) 1,017 s Grok 4.5: Accuracy 71%, latency (mean per test) 848 s GPT-5.6 Luna: Accuracy 73.6%, latency (mean per test) 975 s GPT-5.6 Terra: Accuracy 72.1%, latency (mean per test) 1,024 s Inkling: Accuracy 66.8%, latency (mean per test) 811 s Inkling-Small: Accuracy 64.1%, latency (mean per test) 1,175 s Kimi K3: Accuracy 75.2%, latency (mean per test) 1,844 s Gemini 3.5 Flash-Lite: Accuracy 59.1%, latency (mean per test) 1,076 s Gemini 3.6 Flash: Accuracy 45.4%, latency (mean per test) 660 s Claude Opus 5: Accuracy 65.4%, latency (mean per test) 1,165 s DeepSeek V4 Flash 0731: Accuracy 72.1%, latency (mean per test) 928 s Qwen 3.8 Max: Accuracy 28.6%, latency (mean per test) 580 s Muse Spark 1.2: Accuracy 69.5%, latency (mean per test) 1,221 s Nemotron 3.5 Lightning 30B-A3B: Accuracy 29.5%, latency (mean per test) 3,032 s Grok 4.6: Accuracy 67.7%, latency (mean per test) 2,212 s DeepSeek V4 Pro 0813: Accuracy 72.9%, latency (mean per test) 1,353 s GLM-5.3: Accuracy 72.1%, latency (mean per test) 1,819 s Qwen 3.8 27B: Accuracy 67.8%, latency (mean per test) 2,496 s GLM-5.3-Flash: Accuracy 67.9%, latency (mean per test) 2,148 s Tencent Hy4 preview: Accuracy 65.4%, latency (mean per test) 2,203 s Claude Fable 5.1: Accuracy 70.4%, latency (mean per test) 1,112 s Gemini 3.8 Flash: Accuracy 43.8%, latency (mean per test) 589 s Muse Spark 1.3 (max): Accuracy 72.7%, latency (mean per test) 978 s Muse Spark 1.3: Accuracy 69.4%, latency (mean per test) 1,493 s Ling-3.0-flash-Fin: Accuracy 58.3%, latency (mean per test) 1,163 s Mercury 2.5: Accuracy 45.5%, latency (mean per test) 1,139 s DeepSeek V4.1 Flash: Accuracy 73.7%, latency (mean per test) 830 s Grok 4.7: Accuracy 69.5%, latency (mean per test) 930 s MiMo-V2.6 Flash: Accuracy 75.4%, latency (mean per test) 1,702 s MiMo-V2.6 Pro: Accuracy 72.9%, latency (mean per test) 1,661 s Claude Opus 5.5: Accuracy 55.4%, latency (mean per test) 1,091 s GPT-6 Luna: Accuracy 76.3%, latency (mean per test) 950 s Ember-1: Accuracy 69.5%, latency (mean per test) 1,322 s Claude Sonnet 5.5: Accuracy 59.6%, latency (mean per test) 2,033 s Gemini 4 Argon: Accuracy 77.9%, latency (mean per test) 1,150 s Claude Haiku 5.5: Accuracy 75.5%, latency (mean per test) 1,959 s Muse Wonder Full: Accuracy 73.6%, latency (mean per test) 2,127 s Inkling 2 Small: Accuracy 72.7%, latency (mean per test) 1,455 s Claude Opus 5.5 (CVP): Accuracy 74.6%, latency (mean per test) 1,382 s. Not in our model list Daybreak Blue: Accuracy 73%, latency (mean per test) 710 s. Not in our model list Claude Sonnet 5 (Cyber Range): Accuracy 70.3%, latency (mean per test) 1,377 s. Not in our model list GPT-5.6 Sol: Accuracy 76.3%, latency (mean per test) 647 s Gemini 3.7 Flash: Accuracy 43.8%, latency (mean per test) 454 s GPT-6 Astra: Accuracy 41.1%, latency (mean per test) 250 s GPT-6 Sol: Accuracy 78%, latency (mean per test) 734 s GPT-6.1 Sol: Accuracy 39.3%, latency (mean per test) 222 s Muse Wonder: Accuracy 80.4%, latency (mean per test) 825 s Muse WonderGPT-6 SolGPT-5.6 SolGemini 3.7 FlashGPT-6 AstraGPT-6.1 Sol 0% 20% 40% 60% 80% 100% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy MiniMax-M3: Accuracy 65%, latency (mean per test) 2,456 s Claude Sonnet 5: Accuracy 61.9%, latency (mean per test) 1,017 s Grok 4.5: Accuracy 71%, latency (mean per test) 848 s GPT-5.6 Luna: Accuracy 73.6%, latency (mean per test) 975 s GPT-5.6 Terra: Accuracy 72.1%, latency (mean per test) 1,024 s Inkling: Accuracy 66.8%, latency (mean per test) 811 s Inkling-Small: Accuracy 64.1%, latency (mean per test) 1,175 s Kimi K3: Accuracy 75.2%, latency (mean per test) 1,844 s Gemini 3.5 Flash-Lite: Accuracy 59.1%, latency (mean per test) 1,076 s Gemini 3.6 Flash: Accuracy 45.4%, latency (mean per test) 660 s Claude Opus 5: Accuracy 65.4%, latency (mean per test) 1,165 s DeepSeek V4 Flash 0731: Accuracy 72.1%, latency (mean per test) 928 s Qwen 3.8 Max: Accuracy 28.6%, latency (mean per test) 580 s Muse Spark 1.2: Accuracy 69.5%, latency (mean per test) 1,221 s Nemotron 3.5 Lightning 30B-A3B: Accuracy 29.5%, latency (mean per test) 3,032 s Grok 4.6: Accuracy 67.7%, latency (mean per test) 2,212 s DeepSeek V4 Pro 0813: Accuracy 72.9%, latency (mean per test) 1,353 s GLM-5.3: Accuracy 72.1%, latency (mean per test) 1,819 s Qwen 3.8 27B: Accuracy 67.8%, latency (mean per test) 2,496 s GLM-5.3-Flash: Accuracy 67.9%, latency (mean per test) 2,148 s Tencent Hy4 preview: Accuracy 65.4%, latency (mean per test) 2,203 s Claude Fable 5.1: Accuracy 70.4%, latency (mean per test) 1,112 s Gemini 3.8 Flash: Accuracy 43.8%, latency (mean per test) 589 s Muse Spark 1.3 (max): Accuracy 72.7%, latency (mean per test) 978 s Muse Spark 1.3: Accuracy 69.4%, latency (mean per test) 1,493 s Ling-3.0-flash-Fin: Accuracy 58.3%, latency (mean per test) 1,163 s Mercury 2.5: Accuracy 45.5%, latency (mean per test) 1,139 s DeepSeek V4.1 Flash: Accuracy 73.7%, latency (mean per test) 830 s Grok 4.7: Accuracy 69.5%, latency (mean per test) 930 s MiMo-V2.6 Flash: Accuracy 75.4%, latency (mean per test) 1,702 s MiMo-V2.6 Pro: Accuracy 72.9%, latency (mean per test) 1,661 s Claude Opus 5.5: Accuracy 55.4%, latency (mean per test) 1,091 s GPT-6 Luna: Accuracy 76.3%, latency (mean per test) 950 s Ember-1: Accuracy 69.5%, latency (mean per test) 1,322 s Claude Sonnet 5.5: Accuracy 59.6%, latency (mean per test) 2,033 s Gemini 4 Argon: Accuracy 77.9%, latency (mean per test) 1,150 s Claude Haiku 5.5: Accuracy 75.5%, latency (mean per test) 1,959 s Muse Wonder Full: Accuracy 73.6%, latency (mean per test) 2,127 s Inkling 2 Small: Accuracy 72.7%, latency (mean per test) 1,455 s Claude Opus 5.5 (CVP): Accuracy 74.6%, latency (mean per test) 1,382 s. Not in our model list Daybreak Blue: Accuracy 73%, latency (mean per test) 710 s. Not in our model list Claude Sonnet 5 (Cyber Range): Accuracy 70.3%, latency (mean per test) 1,377 s. Not in our model list GPT-5.6 Sol: Accuracy 76.3%, latency (mean per test) 647 s Gemini 3.7 Flash: Accuracy 43.8%, latency (mean per test) 454 s GPT-6 Astra: Accuracy 41.1%, latency (mean per test) 250 s GPT-6 Sol: Accuracy 78%, latency (mean per test) 734 s GPT-6.1 Sol: Accuracy 39.3%, latency (mean per test) 222 s Muse Wonder: Accuracy 80.4%, latency (mean per test) 825 s Muse WonderGPT-6 SolGPT-5.6 SolGemini 3.7 FlashGPT-6 AstraGPT-6.1 Sol
ModelSettingAccuracyLatency (mean per test)
Muse Wonder 80.4% 824.81 s
GPT-6 Sol 78% 734.45 s
Gemini 4 Argon 77.9% 1,150 s
GPT-5.6 Sol 76.3% 646.76 s
GPT-6 Luna 76.3% 949.8 s
Claude Haiku 5.5 75.5% 1,959 s
MiMo-V2.6 Flash 75.4% 1,702 s
Kimi K3 75.2% 1,844 s
Claude Opus 5.5 (CVP) 74.6% 1,382 s
DeepSeek V4.1 Flash 73.7% 830.11 s
GPT-5.6 Luna 73.6% 975.2 s
Muse Wonder Full 73.6% 2,127 s
Daybreak Blue 73% 710.02 s
DeepSeek V4 Pro 0813 72.9% 1,353 s
MiMo-V2.6 Pro 72.9% 1,661 s
Muse Spark 1.3 max 72.7% 978.43 s
Inkling 2 Small 72.7% 1,455 s
DeepSeek V4 Flash 0731 72.1% 927.82 s
GPT-5.6 Terra 72.1% 1,024 s
GLM-5.3 72.1% 1,819 s
Grok 4.5 71% 848.12 s
Claude Fable 5.1 70.4% 1,112 s
Claude Sonnet 5 (Cyber Range) 70.3% 1,377 s
Muse Spark 1.2 69.5% 1,221 s
Ember-1 69.5% 1,322 s
Grok 4.7 69.5% 930.02 s
Muse Spark 1.3 69.4% 1,493 s
GLM-5.3-Flash 67.9% 2,148 s
Qwen 3.8 27B 67.8% 2,496 s
Grok 4.6 67.7% 2,212 s
Inkling 66.8% 811.11 s
Claude Opus 5 65.4% 1,165 s
Tencent Hy4 preview 65.4% 2,203 s
MiniMax-M3 65% 2,456 s
Inkling-Small 64.1% 1,175 s
Claude Sonnet 5 61.9% 1,017 s
Claude Sonnet 5.5 59.6% 2,033 s
Gemini 3.5 Flash-Lite 59.1% 1,076 s
Ling-3.0-flash-Fin 58.3% 1,163 s
Claude Opus 5.5 55.4% 1,091 s
Mercury 2.5 45.5% 1,139 s
Gemini 3.6 Flash 45.4% 660.04 s
Gemini 3.7 Flash 43.8% 453.76 s
Gemini 3.8 Flash 43.8% 588.85 s
GPT-6 Astra 41.1% 250.41 s
GPT-6.1 Sol 39.3% 221.99 s
Nemotron 3.5 Lightning 30B-A3B 29.5% 3,032 s
Qwen 3.8 Max 28.6% 580.17 s

The best score, 80.4%, costs $3.66 a task

Runs on the cost frontier score higher than any run that's cheaper. MiMo-V2.6 Flash comes within 5 points of the best for $0.05 a task, 73 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Xiaomi Corp OpenAI Meta Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Muse Wonder GPT-6 Sol GPT-6 Luna MiMo-V2.6 Flash 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Muse Wonder GPT-6 Luna

Source: Official CyberBench v1.1 (Vals) leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Muse Wonder 80.4% $3.66 28,669
GPT-6 Sol 78% $1.34 26,196
Gemini 4 Argon 77.9% $5.11 41,303
GPT-5.6 Sol 76.3% $3.05 12,217
GPT-6 Luna 76.3% $0.13 45,968
Claude Haiku 5.5 75.5% $1.25 136,774
MiMo-V2.6 Flash 75.4% $0.05 49,892
Kimi K3 75.2% $2.13 34,131
DeepSeek V4.1 Flash 73.7% $0.11 36,359
GPT-5.6 Luna 73.6% $0.40 27,040
Muse Wonder Full 73.6% $3.63 27,709
DeepSeek V4 Pro 0813 72.9% $0.59 31,216
MiMo-V2.6 Pro 72.9% $0.09 44,404
Muse Spark 1.3 max 72.7% $3.34 49,381
Inkling 2 Small 72.7% $1.27 67,409
GPT-5.6 Terra 72.1% $3.31 20,611
DeepSeek V4 Flash 0731 72.1% $0.19 36,702
GLM-5.3 72.1% $2.69 37,768
Grok 4.5 71% $2.95 19,683
Claude Fable 5.1 70.4% $4.01 35,911
Muse Spark 1.2 69.5% $2.15 50,301
Ember-1 69.5% $2.43 25,943
Grok 4.7 69.5% $8.54 44,055
Muse Spark 1.3 69.4% $4.26 59,447
GLM-5.3-Flash 67.9% $0.28 39,878
Qwen 3.8 27B 67.8% $0.99 66,431
Grok 4.6 67.7% $4.07 103,187
Inkling 66.8% $1.46 32,101
Claude Opus 5 65.4% $3.53 34,727
Tencent Hy4 preview 65.4% $0.46 34,192
MiniMax-M3 65% $3.41 40,492
Inkling-Small 64.1% $0.25 29,134
Claude Sonnet 5 61.9% $1.89 42,509
Claude Sonnet 5.5 59.6% $2.91 86,309
Gemini 3.5 Flash-Lite 59.1% $0.43 23,922
Ling-3.0-flash-Fin 58.3% $0.10 43,164
Claude Opus 5.5 55.4% $2.85 42,526
Mercury 2.5 45.5% $0.12 13,990
Gemini 3.6 Flash 45.4% $1.31 23,917
Gemini 3.7 Flash 43.8% $2.15
Gemini 3.8 Flash 43.8% $1.96 34,876
GPT-6 Astra 41.1% $1.84
GPT-6.1 Sol 39.3% $0.29
Nemotron 3.5 Lightning 30B-A3B 29.5% $0.09 42,126
Qwen 3.8 Max 28.6% $0.78

2 models have held number one on the official leaderboard since its release in October 2026

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

OpenAI Meta
Own leaderboard 2 leaders
Muse Wonder 80.4%

Source: Official CyberBench v1.1 (Vals) 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 Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Muse Wonder GPT-5.6 Sol Grok 4.5 MiniMax-M3 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Muse Wonder Grok 4.5

Sources: Official CyberBench v1.1 (Vals) leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Thinking Machines
DateLaunch postScoreAs the post writes it
Nov 5, 2026 Inkling 2 Small 72.7 CyberBench v1.1

One lab claim next to the leaderboard

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 1 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Inkling 2 Small 72.7% CyberBench v1.1 72.7% Leaderboard Same

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

OpenAI Meta
CyberBench v1.1 (Vals) 4 leaders
Muse Wonder 80.4%
CyberBench (Vals) Passed 85% in Jul 2026, by GPT-5.6 Sol
GPT-5.6 Sol 88.1%
CyberBench v1.1 (Vals) 2 leaders
Muse Wonder 80.4%

Source: Official leaderboards

Replaces CyberBench (Vals)

More in Cybersecurity

All 115 in Cybersecurity
  • ExploitGym In use 869 real-vulnerability exploit-development tasks (userspace, V8, Linux kernel). Success means capturing a flag through the intended vulnerability. GPT-6 Astra 42.4%
  • ExploitBench In use 41 recent V8 JavaScript-engine N-day vulnerabilities. Progress measured by 16 grader-verified capability flags across five tiers. GPT-6 Astra 100%
  • SRE-Bench In use Software reverse engineering without source: 262 binary instances from 19 private programs, each with six graded objectives. GPT-6 Astra 56.9%
  • CyScenarioBench In use Scenario-based long-horizon security challenges in sandboxes, organized in task, scenario and campaign tiers. Evaluation set kept private. Gemini 4 Argon 100%
  • ExploitBench - Internal Port In use Agents develop working exploits for recently disclosed vulnerabilities to limit contamination. Scored as arbitrary code execution success rate. GPT-6 Cyber 42%
  • MiMo Cyber Bench In use Agents do in-house cybersecurity tasks that add to public vulnerability reproduction and exploitation benchmarks. The report gives no task count or grading details. MiMo-V2.6 Pro 81.7%