Benchmark Almanac

Antidote

Claude Fable 5 1,099 leads since June 2026
Not in model cards

Experts such as doctors, lawyers and engineers compare model answers to prompts from their own history. They judge which answer would still help them a month later.

Surge built it as a counterweight to LMArena. The first release covers everyday conversations, and agentic and enterprise editions were announced.

Made by
Surge AI
Released
May 2026
Kind
Overall chat quality
Level
Expert
Scored
Graded by people
Answers
Open-ended
Tasks
Kept private
Strongest model tested
Claude Fable 5.1 #7 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
47
Entries on its own leaderboard, 36 models we track

Claude Fable 5 leads with 1,099 among models released by September 2026

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

47 entries. Scroll sideways for the rest. Bars start at 800.

  1. 1,099 Claude Fable 5
  2. 1,091 Gemini 3.1 Pro
  3. 1,087 Gemini 3.8 Flash
  4. 1,083 Gemini 3.6 Flash
  5. 1,079 Gemini 3.5 Flash
  6. 1,077 Kimi K3
  7. 1,075 Gemini 3.7 Flash
  8. 1,067 GLM-5.3
  9. 1,065 Claude Fable 5.1
  10. 1,055 Claude Opus 5
  11. 1,051 Qwen3.7-Max
  12. 1,045 Kimi K2.6
  13. 1,042 Claude Opus 4.6
  14. 1,042 Claude Opus 4.7
  15. 1,040 GLM-5.2
  16. 1,030 Claude Opus 4.8
  17. 1,028 Muse Spark 1.2
  18. 1,027 GPT-5.6 Sol
  19. 1,022 Hy Hy3
  20. 1,018 DeepSeek-V4-Pro
  21. 1,015 DeepSeek V4 Flash Vision
  22. 1,014 Muse Spark 1.1
  23. 1,013 GPT-5.5
  24. 1,012 Claude Sonnet 4.6
  25. 1,007 DeepSeek V4 Pro (preview)
  26. 1,006 Kimi K2.5
  27. 1,001 Grok 4.20 Beta
  28. 999 Qwen3.5-Plus
  29. 992 Hy Hy4 Preview
  30. 991 Inkling Inkling
  31. 991 Gemini 3.5 Flash-Lite
  32. 985 Qwen 3.8 Max
  33. 973 Grok 4.5
  34. 966 DeepSeek V4 Flash (preview)
  35. 966 Grok 4.6
  36. 954 DeepSeek-V3.2
  37. 950 Grok 4.3
  38. 947 Mistral Large 3
  39. 947 ERNIE 5.1
  40. 944 Qwen3.8 Flash
  41. 932 Claude Haiku 4.5
  42. 931 Muse Glimmer 30B
  43. 929 GPT-5.4 Mini
  44. 893 Gemma 3 12B
  45. 888 Nemotron Lightning 3.5 30B A3B
  46. 863 Ernie 4.5 300B
  47. 819 Nova 2 Pro

Bars start at 800.

  1. 1 Claude Fable 5 Adaptive/High 1,099
  2. 2 Gemini 3.1 Pro High reasoning 1,091
  3. 3 Gemini 3.8 Flash Medium reasoning 1,087
  4. 4 Gemini 3.6 Flash Medium reasoning 1,083
  5. 5 Gemini 3.5 Flash Medium reasoning 1,079
  6. 6 Kimi K3 Max reasoning 1,077
  7. 7 Gemini 3.7 Flash Medium reasoning 1,075
  8. 8 GLM-5.3 Max reasoning 1,067
  9. 9 Claude Fable 5.1 Adaptive/High 1,065
  10. 10 Claude Opus 5 Adaptive/High 1,055
  11. 11 Qwen3.7-Max Thinking on 1,051
  12. 12 Kimi K2.6 Thinking on 1,045
  13. 13 Claude Opus 4.6 High reasoning 1,042
  14. 14 Claude Opus 4.7 High reasoning 1,042
  15. 15 GLM-5.2 Auto reasoning 1,040
  16. 16 Claude Opus 4.8 High reasoning 1,030
  17. 17 Muse Spark 1.2 Auto reasoning 1,028
  18. 18 GPT-5.6 Sol Medium reasoning 1,027
  19. 19 Hy Hy3 No reasoning 1,022
  20. 20 DeepSeek-V4-Pro High reasoning 1,018
  21. 21 DeepSeek V4 Flash Vision experimental 1,015
  22. 22 Muse Spark 1.1 Medium reasoning 1,014
  23. 23 GPT-5.5 Medium reasoning 1,013
  24. 24 Claude Sonnet 4.6 High reasoning 1,012
  25. 25 DeepSeek V4 Pro (preview) High reasoning 1,007
  26. 26 Kimi K2.5 Thinking on 1,006
  27. 27 Grok 4.20 Beta 1,001
  28. 28 Qwen3.5-Plus Thinking on 999
  29. 29 Hy Hy4 Preview High reasoning 992
  30. 30 Inkling Inkling High reasoning 991
  31. 31 Gemini 3.5 Flash-Lite Minimal reasoning 991
  32. 32 Qwen 3.8 Max xHigh reasoning 985
  33. 33 Grok 4.5 High reasoning 973
  34. 34 DeepSeek V4 Flash (preview) High reasoning 966
  35. 35 Grok 4.6 High reasoning 966
  36. 36 DeepSeek-V3.2 No reasoning 954
  37. 37 Grok 4.3 default 950
  38. 38 Mistral Large 3 947
  39. 39 ERNIE 5.1 947
  40. 40 Qwen3.8 Flash xHigh reasoning 944
  41. 41 Claude Haiku 4.5 932
  42. 42 Muse Glimmer 30B High reasoning 931
  43. 43 GPT-5.4 Mini No reasoning 929
  44. 44 Gemma 3 12B 893
  45. 45 Nemotron Lightning 3.5 30B A3B Thinking on 888
  46. 46 Ernie 4.5 300B 863
  47. 47 Nova 2 Pro No reasoning 819

Source: Official Antidote leaderboard

#EntryScoreSettingDateSource
1 Claude Fable 5 1,099 Adaptive/High Jun 2026 Official leaderboard
2 Gemini 3.1 Pro 1,091 High reasoning Feb 2026 Official leaderboard
3 Gemini 3.8 Flash 1,087 Medium reasoning Sep 2026 Official leaderboard
4 Gemini 3.6 Flash 1,083 Medium reasoning Jul 2026 Official leaderboard
5 Gemini 3.5 Flash 1,079 Medium reasoning May 2026 Official leaderboard
6 Kimi K3 1,077 Max reasoning Jul 2026 Official leaderboard
7 Gemini 3.7 Flash 1,075 Medium reasoning Aug 2026 Official leaderboard
8 GLM-5.3 1,067 Max reasoning Aug 2026 Official leaderboard
9 Claude Fable 5.1 1,065 Adaptive/High Sep 2026 Official leaderboard
10 Claude Opus 5 1,055 Adaptive/High Jul 2026 Official leaderboard
11 Qwen3.7-Max 1,051 Thinking on May 2026 Official leaderboard
12 Kimi K2.6 1,045 Thinking on Apr 2026 Official leaderboard
13 Claude Opus 4.6 1,042 High reasoning Feb 2026 Official leaderboard
14 Claude Opus 4.7 1,042 High reasoning Apr 2026 Official leaderboard
15 GLM-5.2 1,040 Auto reasoning Jun 2026 Official leaderboard
16 Claude Opus 4.8 1,030 High reasoning May 2026 Official leaderboard
17 Muse Spark 1.2 1,028 Auto reasoning Aug 2026 Official leaderboard
18 GPT-5.6 Sol 1,027 Medium reasoning Jul 2026 Official leaderboard
19 Hy Hy3 1,022 No reasoning Official leaderboard
20 DeepSeek-V4-Pro 1,018 High reasoning Apr 2026 Official leaderboard
21 DeepSeek V4 Flash Vision 1,015 experimental Official leaderboard
22 Muse Spark 1.1 1,014 Medium reasoning Jul 2026 Official leaderboard
23 GPT-5.5 1,013 Medium reasoning Apr 2026 Official leaderboard
24 Claude Sonnet 4.6 1,012 High reasoning Feb 2026 Official leaderboard
25 DeepSeek V4 Pro (preview) 1,007 High reasoning Official leaderboard
26 Kimi K2.5 1,006 Thinking on Jan 2026 Official leaderboard
27 Grok 4.20 Beta 1,001 Official leaderboard
28 Qwen3.5-Plus 999 Thinking on Feb 2026 Official leaderboard
29 Hy Hy4 Preview 992 High reasoning Official leaderboard
30 Inkling Inkling 991 High reasoning Official leaderboard
31 Gemini 3.5 Flash-Lite 991 Minimal reasoning Jul 2026 Official leaderboard
32 Qwen 3.8 Max 985 xHigh reasoning Aug 2026 Official leaderboard
33 Grok 4.5 973 High reasoning Jul 2026 Official leaderboard
34 DeepSeek V4 Flash (preview) 966 High reasoning Official leaderboard
35 Grok 4.6 966 High reasoning Aug 2026 Official leaderboard
36 DeepSeek-V3.2 954 No reasoning Dec 2025 Official leaderboard
37 Grok 4.3 950 default Apr 2026 Official leaderboard
38 Mistral Large 3 947 Dec 2025 Official leaderboard
39 ERNIE 5.1 947 May 2026 Official leaderboard
40 Qwen3.8 Flash 944 xHigh reasoning Aug 2026 Official leaderboard
41 Claude Haiku 4.5 932 Oct 2025 Official leaderboard
42 Muse Glimmer 30B 931 High reasoning Official leaderboard
43 GPT-5.4 Mini 929 No reasoning Mar 2026 Official leaderboard
44 Gemma 3 12B 893 Mar 2025 Official leaderboard
45 Nemotron Lightning 3.5 30B A3B 888 Thinking on Official leaderboard
46 Ernie 4.5 300B 863 Official leaderboard
47 Nova 2 Pro 819 No reasoning Official leaderboard

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

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

Google DeepMind Anthropic
Own leaderboard Not tested since Sep 2026
Claude Fable 5 1,099

Source: Official Antidote 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
850 900 950 1,000 1,050 1,100 2025 2026 850 900 950 1,000 1,050 1,100 2025 2026

Source: Official Antidote 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 Instructions and chat

All 81 in Instructions and chat
  • IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy. Gemma 4 31B IT 98.9%
  • IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose). Nemotron-Cascade 2 30B-A3B 82.9%
  • MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
  • Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
  • AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics. MAI-Thinking-1 85%
  • IFStruct In use Models generate JSON or YAML that must follow a requested schema and phrasing constraints, without constrained decoding. A response passes only if every structural constraint holds. Gemma 4 31B IT 95.9%