Benchmark Almanac

Vals Index v2.2

Claude Fable 5.2 71.3% leads since November 2026
Claude Fable 5.2 71.3% leads since November 2026
In use

Agents are scored on eight finance, coding, legal, and tax benchmarks, combined with weights from each sector's share of US GDP.

Version 2.2 from 19 October 2026: (8.0 × finance + 5.6 × coding + 1.2 × legal + 0.5 × tax) / 15.3. Finance averages Finance Agent v2.1 and EMB v1.1. Coding averages Terminal-Bench 4, Vibe Code Bench v1.1 and Code Migration v1.1. Legal averages Legal Research and HLAB-Vals-1.1. Tax is Tax Agent Bench (all-pass accuracy). Six of the eight components are private. Earlier versions have their own records.

Made by
Vals AI
Released
October 2026
Kind
Composite work indices
Scored
Graded by a model
Answers
Agent task
Tasks
Public, plus a private test set
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 13th, 2026
37
Models scored on its own leaderboard
2
Launch posts list it, from 2 labs

Claude Fable 5.2 leads with 71.3%

Best score per model from the benchmark's own leaderboard, read on November 13th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

Claude Fable 5.2 leads with 71.3%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

37 models. Scroll sideways for the rest.

  1. 71.3 Claude Fable 5.2
  2. 70.8 Claude Sonnet 5.5
  3. 70.5 Gemini 4 Carbon
  4. 70.1 Claude Opus 5.5
  5. 69.4 Gemini 4 Argon
  6. 68.5 Gemini 4 Hydrogen
  7. 66.8 Muse Wonder Full
  8. 65.0 GPT-6 Astra
  9. 64.9 Muse Wonder
  10. 63.9 GPT-6.1 Sol
  11. 62.5 Kimi K3.1 Pro
  12. 61.3 DeepSeek V4.1 Flash 1017
  13. 60.5 GPT-6.1 Terra
  14. 60.5 Muse Ember
  15. 59.9 GLM-5.3
  16. 59.4 Tencent Hy4
  17. 59.0 Qwen3.8 Max (0902)
  18. 58.7 Claude Haiku 5.5
  19. 58.6 Grok 4.7
  20. 57.8 MiMo-V2.6 Pro
  21. 56.5 Mistral Large 4
  22. 56.1 Inkling 2 Small
  23. 56.0 Gemini 3.8 Flash
  24. 54.8 MiMo-V2.6 Flash
  25. 54.8 Step 5 Preview
  26. 53.8 MiniMax-M3.1-Flash
  27. 53.2 Tencent Hy4 preview
  28. 53.0 DeepSeek V4 Pro 0813
  29. 53.0 Kimi K3
  30. 52.3 GPT-6 Luna
  31. 50.9 GLM-5.3-Flash
  32. 48.6 Muse Spark Open
  33. 43.7 Mercury 3
  34. 43.4 Beam 1.0
  35. 29.8 Nemotron 3.5 Super VL
  36. 29.0 Inkling
  37. 27.6 Inkling-Small
  1. 1 Claude Fable 5.2 71.3%
  2. 2 Claude Sonnet 5.5 70.8%
  3. 3 Gemini 4 Carbon 70.5%
  4. 4 Claude Opus 5.5 70.1%
  5. 5 Gemini 4 Argon 69.4%
  6. 6 Gemini 4 Hydrogen 68.5%
  7. 7 Muse Wonder Full 66.8%
  8. 8 GPT-6 Astra 65%
  9. 9 Muse Wonder 64.9%
  10. 10 GPT-6.1 Sol 63.9%
  11. 11 Kimi K3.1 Pro 62.5%
  12. 12 DeepSeek V4.1 Flash 1017 61.3%
  13. 13 GPT-6.1 Terra 60.5%
  14. 14 Muse Ember 60.5%
  15. 15 GLM-5.3 59.9%
  16. 16 Tencent Hy4 59.4%
  17. 17 Qwen3.8 Max (0902) 59%
  18. 18 Claude Haiku 5.5 58.7%
  19. 19 Grok 4.7 58.6%
  20. 20 MiMo-V2.6 Pro 57.8%
  21. 21 Mistral Large 4 56.5%
  22. 22 Inkling 2 Small 56.1%
  23. 23 Gemini 3.8 Flash 56%
  24. 24 MiMo-V2.6 Flash 54.8%
  25. 25 Step 5 Preview 54.8%
  26. 26 MiniMax-M3.1-Flash 53.8%
  27. 27 Tencent Hy4 preview 53.2%
  28. 28 DeepSeek V4 Pro 0813 53%
  29. 29 Kimi K3 53%
  30. 30 GPT-6 Luna 52.3%
  31. 31 GLM-5.3-Flash 50.9%
  32. 32 Muse Spark Open 48.6%
  33. 33 Mercury 3 43.7%
  34. 34 Beam 1.0 43.4%
  35. 35 Nemotron 3.5 Super VL 29.8%
  36. 36 Inkling 29%
  37. 37 Inkling-Small 27.6%

Source: Official Vals Index v2.2 leaderboard

37 models. Scroll sideways for the rest.

  1. 71.3 Claude Fable 5.2
  2. 70.8 Claude Sonnet 5.5
  3. 70.5 Gemini 4 Carbon
  4. 70.1 Claude Opus 5.5
  5. 69.4 Gemini 4 Argon
  6. 68.5 Gemini 4 Hydrogen
  7. 66.8 Muse Wonder Full
  8. 65.0 GPT-6 Astra
  9. 64.9 Muse Wonder
  10. 63.9 GPT-6.1 Sol
  11. 62.5 Kimi K3.1 Pro
  12. 61.3 DeepSeek V4.1 Flash 1017
  13. 60.5 GPT-6.1 Terra
  14. 60.5 Muse Ember
  15. 59.9 GLM-5.3
  16. 59.4 Tencent Hy4
  17. 59.0 Qwen3.8 Max (0902)
  18. 58.7 Claude Haiku 5.5
  19. 58.6 Grok 4.7
  20. 57.8 MiMo-V2.6 Pro
  21. 56.5 Mistral Large 4
  22. 56.1 Inkling 2 Small
  23. 56.0 Gemini 3.8 Flash
  24. 54.8 MiMo-V2.6 Flash
  25. 54.8 Step 5 Preview
  26. 53.8 MiniMax-M3.1-Flash
  27. 53.2 Tencent Hy4 preview
  28. 53.0 DeepSeek V4 Pro 0813
  29. 53.0 Kimi K3
  30. 52.3 GPT-6 Luna
  31. 50.9 GLM-5.3-Flash
  32. 48.6 Muse Spark Open
  33. 43.7 Mercury 3
  34. 43.4 Beam 1.0
  35. 29.8 Nemotron 3.5 Super VL
  36. 29.0 Inkling
  37. 27.6 Inkling-Small
  1. 1 Claude Fable 5.2 71.3%
  2. 2 Claude Sonnet 5.5 70.8%
  3. 3 Gemini 4 Carbon 70.5%
  4. 4 Claude Opus 5.5 70.1%
  5. 5 Gemini 4 Argon 69.4%
  6. 6 Gemini 4 Hydrogen 68.5%
  7. 7 Muse Wonder Full 66.8%
  8. 8 GPT-6 Astra 65%
  9. 9 Muse Wonder 64.9%
  10. 10 GPT-6.1 Sol 63.9%
  11. 11 Kimi K3.1 Pro 62.5%
  12. 12 DeepSeek V4.1 Flash 1017 61.3%
  13. 13 GPT-6.1 Terra 60.5%
  14. 14 Muse Ember 60.5%
  15. 15 GLM-5.3 59.9%
  16. 16 Tencent Hy4 59.4%
  17. 17 Qwen3.8 Max (0902) 59%
  18. 18 Claude Haiku 5.5 58.7%
  19. 19 Grok 4.7 58.6%
  20. 20 MiMo-V2.6 Pro 57.8%
  21. 21 Mistral Large 4 56.5%
  22. 22 Inkling 2 Small 56.1%
  23. 23 Gemini 3.8 Flash 56%
  24. 24 MiMo-V2.6 Flash 54.8%
  25. 25 Step 5 Preview 54.8%
  26. 26 MiniMax-M3.1-Flash 53.8%
  27. 27 Tencent Hy4 preview 53.2%
  28. 28 DeepSeek V4 Pro 0813 53%
  29. 29 Kimi K3 53%
  30. 30 GPT-6 Luna 52.3%
  31. 31 GLM-5.3-Flash 50.9%
  32. 32 Muse Spark Open 48.6%
  33. 33 Mercury 3 43.7%
  34. 34 Beam 1.0 43.4%
  35. 35 Nemotron 3.5 Super VL 29.8%
  36. 36 Inkling 29%
  37. 37 Inkling-Small 27.6%

Sources: Official Vals Index v2.2 leaderboard and labs’ model cards

#ModelScoreSettingDateSource
1 Claude Fable 5.2 71.3% Nov 2026 Official leaderboard
2 Claude Sonnet 5.5 70.8% Sep 2026 Official leaderboard
3 Gemini 4 Carbon 70.5% Oct 2026 Official leaderboard
4 Claude Opus 5.5 70.1% Sep 2026 Official leaderboard
5 Gemini 4 Argon 69.4% Sep 2026 Official leaderboard
6 Gemini 4 Hydrogen 68.5% Nov 2026 Official leaderboard
7 Muse Wonder Full 66.8% Oct 2026 Official leaderboard
8 GPT-6 Astra 65% Sep 2026 Official leaderboard
9 Muse Wonder 64.9% Oct 2026 Official leaderboard
10 GPT-6.1 Sol 63.9% Sep 2026 Official leaderboard
11 Kimi K3.1 Pro 62.5% Oct 2026 Official leaderboard
12 DeepSeek V4.1 Flash 1017 61.3% Oct 2026 Official leaderboard
13 GPT-6.1 Terra 60.5% Oct 2026 Official leaderboard
14 Muse Ember 60.5% Nov 2026 Official leaderboard
15 GLM-5.3 59.9% Aug 2026 Official leaderboard
16 Tencent Hy4 59.4% Nov 2026 Official leaderboard
17 Qwen3.8 Max (0902) 59% Sep 2026 Official leaderboard
18 Claude Haiku 5.5 58.7% Oct 2026 Official leaderboard
19 Grok 4.7 58.6% Sep 2026 Official leaderboard
20 MiMo-V2.6 Pro 57.8% Sep 2026 Official leaderboard
21 Mistral Large 4 56.5% Oct 2026 Official leaderboard
22 Inkling 2 Small 56.1% Nov 2026 Official leaderboard
23 Gemini 3.8 Flash 56% Sep 2026 Official leaderboard
24 MiMo-V2.6 Flash 54.8% Sep 2026 Official leaderboard
25 Step 5 Preview 54.8% Sep 2026 Official leaderboard
26 MiniMax-M3.1-Flash 53.8% Oct 2026 Official leaderboard
27 Tencent Hy4 preview 53.2% Aug 2026 Official leaderboard
28 DeepSeek V4 Pro 0813 53% Aug 2026 Official leaderboard
29 Kimi K3 53% Jul 2026 Official leaderboard
30 GPT-6 Luna 52.3% Sep 2026 Official leaderboard
31 GLM-5.3-Flash 50.9% Aug 2026 Official leaderboard
32 Muse Spark Open 48.6% Oct 2026 Official leaderboard
33 Mercury 3 43.7% Oct 2026 Official leaderboard
34 Beam 1.0 43.4% Oct 2026 Official leaderboard
35 Nemotron 3.5 Super VL 29.8% Oct 2026 Official leaderboard
36 Inkling 29% Jul 2026 Official leaderboard
37 Inkling-Small 27.6% Jul 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Fable 5.2 71.3% Nov 2026 Official leaderboard
2 Claude Sonnet 5.5 70.8% Sep 2026 Official leaderboard
3 Gemini 4 Carbon 70.5% Oct 2026 Official leaderboard
4 Claude Opus 5.5 70.1% Sep 2026 Official leaderboard
5 Gemini 4 Argon 69.4% Sep 2026 Official leaderboard
6 Gemini 4 Hydrogen 68.5% Nov 2026 Official leaderboard
7 Muse Wonder Full 66.8% Oct 2026 Official leaderboard
8 GPT-6 Astra 65% Sep 2026 Official leaderboard
9 Muse Wonder 64.9% Oct 2026 Official leaderboard
10 GPT-6.1 Sol 63.9% Sep 2026 Official leaderboard
11 Kimi K3.1 Pro 62.5% Oct 2026 Official leaderboard
12 DeepSeek V4.1 Flash 1017 61.3% Oct 2026 Official leaderboard
13 GPT-6.1 Terra 60.5% Oct 2026 Official leaderboard
14 Muse Ember 60.5% Nov 2026 Official leaderboard
15 GLM-5.3 59.9% Aug 2026 Official leaderboard
16 Tencent Hy4 59.4% Nov 2026 Official leaderboard
17 Qwen3.8 Max (0902) 59% Sep 2026 Official leaderboard
18 Claude Haiku 5.5 58.7% Oct 2026 Official leaderboard
19 Grok 4.7 58.6% Sep 2026 Official leaderboard
20 MiMo-V2.6 Pro 57.8% Sep 2026 Official leaderboard
21 Mistral Large 4 56.5% Oct 2026 Official leaderboard
22 Inkling 2 Small 56.1% Nov 2026 Lab model card
23 Gemini 3.8 Flash 56% Sep 2026 Official leaderboard
24 MiMo-V2.6 Flash 54.8% Sep 2026 Official leaderboard
25 Step 5 Preview 54.8% Sep 2026 Official leaderboard
26 MiniMax-M3.1-Flash 53.8% Oct 2026 Official leaderboard
27 Tencent Hy4 preview 53.2% Aug 2026 Official leaderboard
28 DeepSeek V4 Pro 0813 53% Aug 2026 Official leaderboard
29 Kimi K3 53% Jul 2026 Official leaderboard
30 GPT-6 Luna 52.3% Sep 2026 Official leaderboard
31 GLM-5.3-Flash 50.9% Aug 2026 Official leaderboard
32 Muse Spark Open 48.6% Oct 2026 Official leaderboard
33 Mercury 3 43.7% Oct 2026 Official leaderboard
34 Beam 1.0 43.4% Oct 2026 Official leaderboard
35 Nemotron 3.5 Super VL 29.8% Oct 2026 Official leaderboard
36 Inkling 29% Jul 2026 Official leaderboard
37 Inkling-Small 27.6% Jul 2026 Official leaderboard

Labs report scores for 2 models in their cards. Switch the scores above to rank them together.

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% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy Inkling: Accuracy 29%, latency (mean per test) 1,033 s Inkling-Small: Accuracy 27.6%, latency (mean per test) 986 s Kimi K3: Accuracy 53%, latency (mean per test) 4,496 s DeepSeek V4 Pro 0813: Accuracy 53%, latency (mean per test) 3,483 s GLM-5.3: Accuracy 59.9%, latency (mean per test) 3,864 s GLM-5.3-Flash: Accuracy 50.9%, latency (mean per test) 3,831 s Tencent Hy4 preview: Accuracy 53.2%, latency (mean per test) 3,563 s Qwen3.8 Max (0902): Accuracy 59%, latency (mean per test) 6,346 s Gemini 3.8 Flash: Accuracy 56%, latency (mean per test) 2,159 s GPT-6 Astra: Accuracy 65%, latency (mean per test) 2,019 s Step 5 Preview: Accuracy 54.8%, latency (mean per test) 3,929 s Grok 4.7: Accuracy 58.6%, latency (mean per test) 2,014 s MiMo-V2.6 Flash: Accuracy 54.8%, latency (mean per test) 4,028 s MiMo-V2.6 Pro: Accuracy 57.8%, latency (mean per test) 4,356 s Claude Opus 5.5: Accuracy 70.1%, latency (mean per test) 5,071 s GPT-6 Luna: Accuracy 52.3%, latency (mean per test) 1,943 s Claude Sonnet 5.5: Accuracy 70.8%, latency (mean per test) 5,159 s GPT-6.1 Sol: Accuracy 63.9%, latency (mean per test) 2,664 s Gemini 4 Argon: Accuracy 69.4%, latency (mean per test) 2,698 s Mistral Large 4: Accuracy 56.5%, latency (mean per test) 2,868 s Claude Haiku 5.5: Accuracy 58.7%, latency (mean per test) 2,997 s Muse Wonder: Accuracy 64.9%, latency (mean per test) 2,124 s Muse Wonder Full: Accuracy 66.8%, latency (mean per test) 3,559 s Muse Spark Open: Accuracy 48.6%, latency (mean per test) 4,925 s Nemotron 3.5 Super VL: Accuracy 29.8%, latency (mean per test) 514 s DeepSeek V4.1 Flash 1017: Accuracy 61.3%, latency (mean per test) 2,395 s Mercury 3: Accuracy 43.7%, latency (mean per test) 1,326 s MiniMax-M3.1-Flash: Accuracy 53.8%, latency (mean per test) 2,598 s Kimi K3.1 Pro: Accuracy 62.5%, latency (mean per test) 3,817 s GPT-6.1 Terra: Accuracy 60.5%, latency (mean per test) 2,383 s Gemini 4 Carbon: Accuracy 70.5%, latency (mean per test) 2,865 s Tencent Hy4: Accuracy 59.4%, latency (mean per test) 3,092 s Inkling 2 Small: Accuracy 56.1%, latency (mean per test) 3,035 s Beam 1.0: Accuracy 43.4%, latency (mean per test) 447 s Claude Fable 5.2: Accuracy 71.3%, latency (mean per test) 2,622 s Muse Ember: Accuracy 60.5%, latency (mean per test) 902 s Gemini 4 Hydrogen: Accuracy 68.5%, latency (mean per test) 1,925 s Claude Fable 5.2Gemini 4 HydrogenMuse EmberBeam 1.0 0% 20% 40% 60% 80% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy Inkling: Accuracy 29%, latency (mean per test) 1,033 s Inkling-Small: Accuracy 27.6%, latency (mean per test) 986 s Kimi K3: Accuracy 53%, latency (mean per test) 4,496 s DeepSeek V4 Pro 0813: Accuracy 53%, latency (mean per test) 3,483 s GLM-5.3: Accuracy 59.9%, latency (mean per test) 3,864 s GLM-5.3-Flash: Accuracy 50.9%, latency (mean per test) 3,831 s Tencent Hy4 preview: Accuracy 53.2%, latency (mean per test) 3,563 s Qwen3.8 Max (0902): Accuracy 59%, latency (mean per test) 6,346 s Gemini 3.8 Flash: Accuracy 56%, latency (mean per test) 2,159 s GPT-6 Astra: Accuracy 65%, latency (mean per test) 2,019 s Step 5 Preview: Accuracy 54.8%, latency (mean per test) 3,929 s Grok 4.7: Accuracy 58.6%, latency (mean per test) 2,014 s MiMo-V2.6 Flash: Accuracy 54.8%, latency (mean per test) 4,028 s MiMo-V2.6 Pro: Accuracy 57.8%, latency (mean per test) 4,356 s Claude Opus 5.5: Accuracy 70.1%, latency (mean per test) 5,071 s GPT-6 Luna: Accuracy 52.3%, latency (mean per test) 1,943 s Claude Sonnet 5.5: Accuracy 70.8%, latency (mean per test) 5,159 s GPT-6.1 Sol: Accuracy 63.9%, latency (mean per test) 2,664 s Gemini 4 Argon: Accuracy 69.4%, latency (mean per test) 2,698 s Mistral Large 4: Accuracy 56.5%, latency (mean per test) 2,868 s Claude Haiku 5.5: Accuracy 58.7%, latency (mean per test) 2,997 s Muse Wonder: Accuracy 64.9%, latency (mean per test) 2,124 s Muse Wonder Full: Accuracy 66.8%, latency (mean per test) 3,559 s Muse Spark Open: Accuracy 48.6%, latency (mean per test) 4,925 s Nemotron 3.5 Super VL: Accuracy 29.8%, latency (mean per test) 514 s DeepSeek V4.1 Flash 1017: Accuracy 61.3%, latency (mean per test) 2,395 s Mercury 3: Accuracy 43.7%, latency (mean per test) 1,326 s MiniMax-M3.1-Flash: Accuracy 53.8%, latency (mean per test) 2,598 s Kimi K3.1 Pro: Accuracy 62.5%, latency (mean per test) 3,817 s GPT-6.1 Terra: Accuracy 60.5%, latency (mean per test) 2,383 s Gemini 4 Carbon: Accuracy 70.5%, latency (mean per test) 2,865 s Tencent Hy4: Accuracy 59.4%, latency (mean per test) 3,092 s Inkling 2 Small: Accuracy 56.1%, latency (mean per test) 3,035 s Beam 1.0: Accuracy 43.4%, latency (mean per test) 447 s Claude Fable 5.2: Accuracy 71.3%, latency (mean per test) 2,622 s Muse Ember: Accuracy 60.5%, latency (mean per test) 902 s Gemini 4 Hydrogen: Accuracy 68.5%, latency (mean per test) 1,925 s Claude Fable 5.2Gemini 4 HydrogenMuse EmberBeam 1.0
ModelSettingAccuracyLatency (mean per test)
Claude Fable 5.2 71.3% 2,622 s
Claude Sonnet 5.5 70.9% 5,159 s
Gemini 4 Carbon 70.5% 2,865 s
Claude Opus 5.5 70.1% 5,071 s
Gemini 4 Argon 69.4% 2,698 s
Gemini 4 Hydrogen 68.5% 1,925 s
Muse Wonder Full 66.8% 3,559 s
GPT-6 Astra 65% 2,019 s
Muse Wonder 64.9% 2,124 s
GPT-6.1 Sol 63.9% 2,664 s
Kimi K3.1 Pro 62.5% 3,817 s
DeepSeek V4.1 Flash 1017 61.3% 2,395 s
GPT-6.1 Terra 60.5% 2,383 s
Muse Ember 60.5% 902.37 s
GLM-5.3 59.9% 3,864 s
Tencent Hy4 59.4% 3,092 s
Qwen3.8 Max (0902) 59.1% 6,346 s
Claude Haiku 5.5 58.7% 2,997 s
Grok 4.7 58.6% 2,014 s
MiMo-V2.6 Pro 57.8% 4,356 s
Mistral Large 4 56.5% 2,868 s
Inkling 2 Small 56.1% 3,035 s
Gemini 3.8 Flash 56% 2,159 s
MiMo-V2.6 Flash 54.8% 4,028 s
Step 5 Preview 54.8% 3,929 s
MiniMax-M3.1-Flash 53.8% 2,598 s
Tencent Hy4 preview 53.2% 3,563 s
DeepSeek V4 Pro 0813 53% 3,483 s
Kimi K3 53% 4,496 s
GPT-6 Luna 52.3% 1,943 s
GLM-5.3-Flash 50.9% 3,831 s
Muse Spark Open 48.6% 4,925 s
Mercury 3 43.7% 1,326 s
Beam 1.0 43.4% 446.56 s
Nemotron 3.5 Super VL 29.8% 514.01 s
Inkling 29% 1,033 s
Inkling-Small 27.6% 986.44 s

The best score, 71.3%, costs $18 a task

Runs on the cost frontier score higher than any run that's cheaper. Muse Wonder Full comes within 4.5 points of the best for $6.81 a task, 3 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Anthropic OpenAI Meta DeepSeek Google DeepMind Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5.2 Muse Wonder Full DeepSeek V4.1 Flash 1017 0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5.2 GPT-6.1 Sol

Source: Official Vals Index v2.2 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 71.3% $18 197,828
Claude Sonnet 5.5 70.8% $16 594,550
Gemini 4 Carbon 70.5% $11 205,620
Claude Opus 5.5 70.1% $25 455,040
Gemini 4 Argon 69.4% $13 159,229
Gemini 4 Hydrogen 68.5% $9.84 336,849
Muse Wonder Full 66.8% $6.81 103,433
GPT-6 Astra 65% $21 99,102
Muse Wonder 64.9% $11 115,812
GPT-6.1 Sol 63.9% $3.61 112,502
Kimi K3.1 Pro 62.5% $6.44 159,072
DeepSeek V4.1 Flash 1017 61.3% $0.66 258,045
GPT-6.1 Terra 60.5% $2.94 157,426
Muse Ember 60.5% $0.15 79,013
GLM-5.3 59.9% $8.52 202,266
Tencent Hy4 59.4% $1.72 238,203
Qwen3.8 Max (0902) 59% $8.73 231,724
Claude Haiku 5.5 58.7% $3.17 570,127
Grok 4.7 58.6% $12 124,146
MiMo-V2.6 Pro 57.8% $0.43 147,277
Mistral Large 4 56.5% $3.72 122,413
Inkling 2 Small 56.1% $3.51 281,311
Gemini 3.8 Flash 56% $5.03 130,206
MiMo-V2.6 Flash 54.8% $0.22 172,167
Step 5 Preview 54.8% $2.73 159,796
MiniMax-M3.1-Flash 53.8% $0.66 255,285
Tencent Hy4 preview 53.2% $1.57 160,621
DeepSeek V4 Pro 0813 53% $2.50 169,366
Kimi K3 53% $7.17 107,342
GPT-6 Luna 52.3% $0.45 170,502
GLM-5.3-Flash 50.9% $2.05 203,647
Muse Spark Open 48.6% $4.72 65,357
Mercury 3 43.7% $0.48 117,494
Beam 1.0 43.4% 41,553
Nemotron 3.5 Super VL 29.8% $0.63 96,498
Inkling 29% $1.85 64,607
Inkling-Small 27.6% $0.40 55,978

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

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

Anthropic
Own leaderboard 2 leaders
Claude Fable 5.2 71.3%

Source: Official Vals Index v2.2 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% 2027 Near the ceiling (85%) Claude Fable 5.2 Claude Sonnet 5.5 GPT-6 Astra GLM-5.3 DeepSeek V4 Pro 0813 Inkling 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Claude Fable 5.2 GPT-6 Astra

Sources: Official Vals Index v2.2 leaderboard and labs’ model cards

2 labs 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.

Google DeepMind
Thinking Machines
DateLaunch postScoreAs the post writes it
Nov 5, 2026 Inkling 2 Small 56.1 Vals Index (v2.2)
Oct 28, 2026 Gemini 4 Carbon 70.4 Vals Index v2.2

2 lab claims 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. 2 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Gemini 4 Carbon 70.4% Vals Index v2.2 70.5% Leaderboard Same
Inkling 2 Small 56.1% Vals Index (v2.2) 56.1% 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.

Anthropic OpenAI Google DeepMind
Vals Index v2.2 16 leaders
Claude Fable 5.2 71.3%
Vals Index v1 Replaced by 2, Aug 2026
Claude Opus 4.7 72.2%
Vals Index v2 Replaced by 2.1, Sep 2026
Claude Opus 5.5 69.7%
Vals Index v2.1 4 leaders
Gemini 4 Argon 68.9%
Vals Index v2.2 2 leaders
Claude Fable 5.2 71.3%

Source: Official leaderboards

Replaces Vals Index v2.1

More in Economic work

All 97 in Economic work
  • JobBench In use 130 tasks from 35 occupations workers most want delegated. Agent works through messy reference files, graded by fact-anchored binary rubrics. GLM-5.3 61.4%
  • GDPval In use Real work tasks from 44 occupations with reference files. Experts blindly compare model deliverables with expert work. Metric is win-or-tie rate. Seed2.1 Pro 87.9%
  • GDPval-AA v2.1 In use Agents with shell and web access complete 220 GDPval gold tasks from 44 occupations. Blind pairwise comparisons give Elo ratings. Claude Opus 5.5 1,846
  • AA-Briefcase v1.1 In use Version 1.1 of Artificial Analysis's multi-hour office work evaluation, with Elo ratings fitted by Crowd-BT. Claude Opus 5.5 1,822
  • CoWorkBench In use Agent completes long-horizon productivity tasks across computer science, finance, law and medicine. Qwen in-house scoring (method unpublished). Claude Fable 5 75.9%
  • OfficeQA In use Grounded reasoning over a large corpus of U.S. Treasury Bulletin documents: locate tables and compute precise numeric answers. Claude Opus 4.7 86.3%