Benchmark Almanac

Vals Finance Index

Gemini 4 Neon 68.3% leads since November 2026
Gemini 4 Neon 68.3% leads since November 2026
In use

The official leaderboard hasn’t scored a model released after Gemini 4 Neon (November 2026). Labs report their own numbers for one newer model: Inkling 2 Mini at 50.1%.

Agents are scored on Finance Agent v2.1, Excel Modeling Benchmark v1.1 and Tax Agent Bench, combined with weights from each sub-sector share of US GDP.

Securities work (1.6% of GDP) averages Finance Agent and EMB. Tax work (0.5%) uses Tax Agent Bench all-pass accuracy. 35 models on October 29th, 2026.

Made by
Vals AI
Released
October 2026
Kind
Composite work indices
Level
Expert
Scored
Graded by a model
Answers
Agent task
Tasks
Kept private
Measures
Capability
Language
English
In and out
Text, Documents, Tables → Actions, Text, Tables
Interaction
Agentic
Items from
Another benchmark
Refresh
One fixed set
Run by
A host
Strongest model tested
GPT-6.2 Astra #1 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 19th, 2026
45
Entries on its own leaderboard, 44 models we track
2
Launch posts list it, from 1 lab
1
System card reports it, from 1 lab

Gemini 4 Neon leads with 68.3% Gemini 4 Neon leads with 68.3%

Best score per entry from the benchmark's own leaderboard, read on November 19th, 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. This view shows only models we track, each at its best.

Gemini 4 Neon leads with 68.3%, 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. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.

  1. 1 Gemini 4 Neon 68.3%
  2. 2 Gemini 4 Carbon 67%
  3. 3 Gemini 4 Hydrogen 66.1%
  4. 4 Gemini 4 Argon 66%
  5. 5 Claude Opus 5.5 65.3%
  6. 6 Muse Wonder Full 64.5%
  7. 7 Claude Sonnet 5.5 64.5%
  8. 8 Claude Fable 5.2 63.6%
  9. 9 Qwen4-Max 62%
  10. 10 Qwen3.8 Max (0902) 61.7%
  11. 11 Inkling 2 Small 61.5%
  12. 12 GLM-5.3 61.4%
  13. 13 Muse Wonder 61.2%
  14. 14 Muse Ember 60.9%
  15. 15 DeepSeek V4.1 Flash 1017 60.9%
  16. 16 Gemini 3.8 Flash 60.5%
  17. 17 Grok 4.7 58.8%
  18. 18 Tencent Hy4 58%
  19. 19 Kimi K3 58%
  20. 20 MiMo-V2.6 Pro 57.9%
  21. 21 Tencent Hy4 preview 57.5%
  22. 22 Claude Haiku 5.5 57.5%
  23. 23 GPT-6.2 Astra 57.3%
  24. 24 MiMo-V2.6 Flash 57.1%
  25. 25 DeepSeek V4.1 Flash 57.1%
  26. 26 Mistral Large 4 57%
  27. 27 Mistral Large 4 Preview 56.9%
  28. 28 GPT-6.2 Sol 56.7%
  29. 29 Kimi K3.1 Pro 56%
  30. 30 MiniMax-M3.1-Flash 55.9%
  31. 31 GLM-5.3-Flash 55.8%
  32. 32 GPT-6.1 Sol 55.8%
  33. 33 GPT-6 Astra 55.6%
  34. 34 Step 5 Preview 55.4%
  35. 35 GPT-6.1 Terra 53.3%
  36. 36 GPT-6 Luna 53%
  37. 37 Muse Spark Open 53%
  38. 38 DeepSeek V4 Pro 0813 52.9%
  39. 39 Hy4 Flash 52.7%
  40. 40 MiniMax-M3 48.4%
  41. 41 Mercury 3 47.2%
  42. 42 Beam 1.0 46.5%
  43. 43 Nemotron 3.5 Super VL 41%
  44. 44 Inkling 38.3%
  45. 45 Inkling-Small 36.3%

Source: Official Vals Finance Index leaderboard

Source: Official Vals Finance Index leaderboard

Sources: Official Vals Finance Index leaderboard and labs’ model cards

#EntryScore ± 95% CISettingDateSource
1 Gemini 4 Neon 68.3% ±2.7 Nov 2026 Official leaderboard
2 Gemini 4 Carbon 67% ±2.8 Oct 2026 Official leaderboard
3 Gemini 4 Hydrogen 66.1% ±2.7 Nov 2026 Official leaderboard
4 Gemini 4 Argon 66% ±2.9 Sep 2026 Official leaderboard
5 Claude Opus 5.5 65.3% ±2.9 Sep 2026 Official leaderboard
6 Muse Wonder Full 64.5% ±2.7 Oct 2026 Official leaderboard
7 Claude Sonnet 5.5 64.5% ±2.9 Sep 2026 Official leaderboard
8 Claude Fable 5.2 63.6% ±2.7 Nov 2026 Official leaderboard
9 Qwen4-Max 62% ±2.8 Nov 2026 Official leaderboard
10 Qwen3.8 Max (0902) 61.7% ±2.9 Sep 2026 Official leaderboard
11 Inkling 2 Small 61.5% ±2.8 Nov 2026 Official leaderboard
12 GLM-5.3 61.4% ±3 Aug 2026 Official leaderboard
13 Muse Wonder 61.2% ±2.9 Oct 2026 Official leaderboard
14 Muse Ember 60.9% ±2.8 Nov 2026 Official leaderboard
15 DeepSeek V4.1 Flash 1017 60.9% ±2.9 Oct 2026 Official leaderboard
16 Gemini 3.8 Flash 60.5% ±2.8 Sep 2026 Official leaderboard
17 Grok 4.7 58.8% ±2.9 Sep 2026 Official leaderboard
18 Tencent Hy4 58% ±2.9 Nov 2026 Official leaderboard
19 Kimi K3 58% ±2.8 Jul 2026 Official leaderboard
20 MiMo-V2.6 Pro 57.9% ±3.1 Sep 2026 Official leaderboard
21 Tencent Hy4 preview 57.5% ±2.8 Aug 2026 Official leaderboard
22 Claude Haiku 5.5 57.5% ±3.5 Oct 2026 Official leaderboard
23 GPT-6.2 Astra 57.3% ±2.7 Nov 2026 Official leaderboard
24 MiMo-V2.6 Flash 57.1% ±2.7 Sep 2026 Official leaderboard
25 DeepSeek V4.1 Flash 57.1% ±2.8 Sep 2026 Official leaderboard
26 Mistral Large 4 57% ±2.8 Oct 2026 Official leaderboard
27 Mistral Large 4 Preview 56.9% ±2.8 Oct 2026 Official leaderboard
28 GPT-6.2 Sol 56.7% ±2.7 Nov 2026 Official leaderboard
29 Kimi K3.1 Pro 56% ±2.9 Oct 2026 Official leaderboard
30 MiniMax-M3.1-Flash 55.9% ±2.8 Oct 2026 Official leaderboard
31 GLM-5.3-Flash 55.8% ±3 Aug 2026 Official leaderboard
32 GPT-6.1 Sol 55.8% ±2.9 Sep 2026 Official leaderboard
33 GPT-6 Astra 55.6% ±2.7 Sep 2026 Official leaderboard
34 Step 5 Preview 55.4% ±2.8 Sep 2026 Official leaderboard
35 GPT-6.1 Terra 53.3% ±2.6 Oct 2026 Official leaderboard
36 GPT-6 Luna 53% ±2.8 Sep 2026 Official leaderboard
37 Muse Spark Open 53% ±2.8 Oct 2026 Official leaderboard
38 DeepSeek V4 Pro 0813 52.9% ±2.8 Aug 2026 Official leaderboard
39 Hy4 Flash 52.7% ±2.7 Official leaderboard
40 MiniMax-M3 48.4% ±2.8 Jun 2026 Official leaderboard
41 Mercury 3 47.2% ±2.9 Oct 2026 Official leaderboard
42 Beam 1.0 46.5% ±2.5 Oct 2026 Official leaderboard
43 Nemotron 3.5 Super VL 41% ±2.8 Oct 2026 Official leaderboard
44 Inkling 38.3% ±3 Jul 2026 Official leaderboard
45 Inkling-Small 36.3% ±3.1 Jul 2026 Official leaderboard
#ModelScore ± 95% CISettingDateSource
1 Gemini 4 Neon 68.3% ±2.7 Nov 2026 Official leaderboard
2 Gemini 4 Carbon 67% ±2.8 Oct 2026 Official leaderboard
3 Gemini 4 Hydrogen 66.1% ±2.7 Nov 2026 Official leaderboard
4 Gemini 4 Argon 66% ±2.9 Sep 2026 Official leaderboard
5 Claude Opus 5.5 65.3% ±2.9 Sep 2026 Official leaderboard
6 Muse Wonder Full 64.5% ±2.7 Oct 2026 Official leaderboard
7 Claude Sonnet 5.5 64.5% ±2.9 Sep 2026 Official leaderboard
8 Claude Fable 5.2 63.6% ±2.7 Nov 2026 Official leaderboard
9 Qwen4-Max 62% ±2.8 Nov 2026 Official leaderboard
10 Qwen3.8 Max (0902) 61.7% ±2.9 Sep 2026 Official leaderboard
11 Inkling 2 Small 61.5% ±2.8 Nov 2026 Official leaderboard
12 GLM-5.3 61.4% ±3 Aug 2026 Official leaderboard
13 Muse Wonder 61.2% ±2.9 Oct 2026 Official leaderboard
14 Muse Ember 60.9% ±2.8 Nov 2026 Official leaderboard
15 DeepSeek V4.1 Flash 1017 60.9% ±2.9 Oct 2026 Official leaderboard
16 Gemini 3.8 Flash 60.5% ±2.8 Sep 2026 Official leaderboard
17 Grok 4.7 58.8% ±2.9 Sep 2026 Official leaderboard
18 Tencent Hy4 58% ±2.9 Nov 2026 Official leaderboard
19 Kimi K3 58% ±2.8 Jul 2026 Official leaderboard
20 MiMo-V2.6 Pro 57.9% ±3.1 Sep 2026 Official leaderboard
21 Tencent Hy4 preview 57.5% ±2.8 Aug 2026 Official leaderboard
22 Claude Haiku 5.5 57.5% ±3.5 Oct 2026 Official leaderboard
23 GPT-6.2 Astra 57.3% ±2.7 Nov 2026 Official leaderboard
24 MiMo-V2.6 Flash 57.1% ±2.7 Sep 2026 Official leaderboard
25 DeepSeek V4.1 Flash 57.1% ±2.8 Sep 2026 Official leaderboard
26 Mistral Large 4 57% ±2.8 Oct 2026 Official leaderboard
27 Mistral Large 4 Preview 56.9% ±2.8 Oct 2026 Official leaderboard
28 GPT-6.2 Sol 56.7% ±2.7 Nov 2026 Official leaderboard
29 Kimi K3.1 Pro 56% ±2.9 Oct 2026 Official leaderboard
30 MiniMax-M3.1-Flash 55.9% ±2.8 Oct 2026 Official leaderboard
31 GLM-5.3-Flash 55.8% ±3 Aug 2026 Official leaderboard
32 GPT-6.1 Sol 55.8% ±2.9 Sep 2026 Official leaderboard
33 GPT-6 Astra 55.6% ±2.7 Sep 2026 Official leaderboard
34 Step 5 Preview 55.4% ±2.8 Sep 2026 Official leaderboard
35 GPT-6.1 Terra 53.3% ±2.6 Oct 2026 Official leaderboard
36 GPT-6 Luna 53% ±2.8 Sep 2026 Official leaderboard
37 Muse Spark Open 53% ±2.8 Oct 2026 Official leaderboard
38 DeepSeek V4 Pro 0813 52.9% ±2.8 Aug 2026 Official leaderboard
39 MiniMax-M3 48.4% ±2.8 Jun 2026 Official leaderboard
40 Mercury 3 47.2% ±2.9 Oct 2026 Official leaderboard
41 Beam 1.0 46.5% ±2.5 Oct 2026 Official leaderboard
42 Nemotron 3.5 Super VL 41% ±2.8 Oct 2026 Official leaderboard
43 Inkling 38.3% ±3 Jul 2026 Official leaderboard
44 Inkling-Small 36.3% ±3.1 Jul 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 4 Neon 68.3% Nov 2026 Official leaderboard
2 Gemini 4 Carbon 67% Oct 2026 Official leaderboard
3 Gemini 4 Hydrogen 66.1% Nov 2026 Official leaderboard
4 Gemini 4 Argon 66% Sep 2026 Official leaderboard
5 Claude Opus 5.5 65.3% Sep 2026 Official leaderboard
6 Muse Wonder Full 64.5% Oct 2026 Official leaderboard
7 Claude Sonnet 5.5 64.5% Sep 2026 Official leaderboard
8 Claude Fable 5.2 63.6% Nov 2026 Official leaderboard
9 Qwen4-Max 62% Nov 2026 Official leaderboard
10 Qwen3.8 Max (0902) 61.7% Sep 2026 Official leaderboard
11 Inkling 2 Small 61.5% Nov 2026 Lab model card
12 GLM-5.3 61.4% Aug 2026 Official leaderboard
13 Muse Wonder 61.2% Oct 2026 Official leaderboard
14 Muse Ember 60.9% Nov 2026 Official leaderboard
15 DeepSeek V4.1 Flash 1017 60.9% Oct 2026 Official leaderboard
16 Gemini 3.8 Flash 60.5% Sep 2026 Official leaderboard
17 Grok 4.7 58.8% Sep 2026 Official leaderboard
18 Tencent Hy4 58% Nov 2026 Official leaderboard
19 Kimi K3 58% Jul 2026 Official leaderboard
20 MiMo-V2.6 Pro 57.9% Sep 2026 Official leaderboard
21 Tencent Hy4 preview 57.5% Aug 2026 Official leaderboard
22 Claude Haiku 5.5 57.5% Oct 2026 Official leaderboard
23 GPT-6.2 Astra 57.3% Nov 2026 Official leaderboard
24 MiMo-V2.6 Flash 57.1% Sep 2026 Official leaderboard
25 DeepSeek V4.1 Flash 57.1% Sep 2026 Official leaderboard
26 Mistral Large 4 57% Oct 2026 Official leaderboard
27 Mistral Large 4 Preview 56.9% Oct 2026 Official leaderboard
28 GPT-6.2 Sol 56.7% Nov 2026 Official leaderboard
29 Kimi K3.1 Pro 56% Oct 2026 Official leaderboard
30 MiniMax-M3.1-Flash 55.9% Oct 2026 Official leaderboard
31 GLM-5.3-Flash 55.8% Aug 2026 Official leaderboard
32 GPT-6.1 Sol 55.8% Sep 2026 Official leaderboard
33 GPT-6 Astra 55.6% Sep 2026 Official leaderboard
34 Step 5 Preview 55.4% Sep 2026 Official leaderboard
35 GPT-6.1 Terra 53.3% Oct 2026 Official leaderboard
36 GPT-6 Luna 53% Sep 2026 Official leaderboard
37 Muse Spark Open 53% Oct 2026 Official leaderboard
38 DeepSeek V4 Pro 0813 52.9% Aug 2026 Official leaderboard
39 Inkling 2 Mini 50.1% Nov 2026 Lab model card
40 MiniMax-M3 48.4% Jun 2026 Official leaderboard
41 Mercury 3 47.2% Oct 2026 Official leaderboard
42 Beam 1.0 46.5% Oct 2026 Official leaderboard
43 Nemotron 3.5 Super VL 41% Oct 2026 Official leaderboard
44 Inkling 38.3% Jul 2026 Official leaderboard
45 Inkling-Small 36.3% 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.

40% 50% 60% 70% 80% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy Gemini 4 NeonGemini 4 HydrogenClaude Fable 5.2Muse Ember 40% 50% 60% 70% 80% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy Gemini 4 NeonGemini 4 HydrogenClaude Fable 5.2Muse Ember
ModelSettingAccuracyLatency (mean per test)
Gemini 4 Neon 68.3% 1,195 s
Gemini 4 Carbon 67.1% 1,838 s
Gemini 4 Hydrogen 66.1% 945.17 s
Gemini 4 Argon 66% 1,362 s
Claude Opus 5.5 65.3% 2,700 s
Muse Wonder Full 64.6% 1,206 s
Claude Sonnet 5.5 64.5% 2,611 s
Claude Fable 5.2 63.6% 495.48 s
Qwen4-Max 62% 1,303 s
Qwen3.8 Max (0902) 61.7% 3,915 s
Inkling 2 Small 61.5% 514.76 s
GLM-5.3 61.4% 2,144 s
Muse Wonder 61.2% 612.47 s
Muse Ember 60.9% 156.44 s
DeepSeek V4.1 Flash 1017 60.9% 1,382 s
Gemini 3.8 Flash 60.5% 463.81 s
Grok 4.7 58.8% 1,249 s
Tencent Hy4 58.1% 2,369 s
Kimi K3 58% 1,344 s
MiMo-V2.6 Pro 57.9% 1,848 s
Tencent Hy4 preview 57.5% 1,935 s
Claude Haiku 5.5 57.5% 2,011 s
GPT-6.2 Astra 57.3% 2,412 s
MiMo-V2.6 Flash 57.2% 1,166 s
DeepSeek V4.1 Flash 57.1% 696.59 s
Mistral Large 4 57% 932.62 s
Mistral Large 4 Preview 56.9% 3,110 s
GPT-6.2 Sol 56.7% 1,607 s
Kimi K3.1 Pro 56% 2,300 s
MiniMax-M3.1-Flash 55.9% 1,572 s
GLM-5.3-Flash 55.8% 2,029 s
GPT-6.1 Sol 55.8% 1,156 s
GPT-6 Astra 55.6% 977.87 s
Step 5 Preview 55.4% 1,482 s
GPT-6.1 Terra 53.3% 1,765 s
GPT-6 Luna 53.1% 1,321 s
Muse Spark Open 53% 2,464 s
DeepSeek V4 Pro 0813 52.9% 1,440 s
Hy4 Flash 52.7% 381.34 s
MiniMax-M3 48.4% 600.9 s
Mercury 3 47.2% 344.44 s
Beam 1.0 46.5% 449.23 s
Nemotron 3.5 Super VL 41% 322.96 s
Inkling 38.3% 988.71 s
Inkling-Small 36.3% 624.36 s

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.

40% 50% 60% 70% 80% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy Gemini 4 NeonGemini 4 HydrogenClaude Fable 5.2Muse Ember 40% 50% 60% 70% 80% 1001,00010K Latency (mean per test) (s), log scale, lower is better Accuracy Gemini 4 NeonGemini 4 HydrogenClaude Fable 5.2Muse Ember
ModelSettingAccuracyLatency (mean per test)
Gemini 4 Neon 68.3% 1,195 s
Gemini 4 Carbon 67.1% 1,838 s
Gemini 4 Hydrogen 66.1% 945.17 s
Gemini 4 Argon 66% 1,362 s
Claude Opus 5.5 65.3% 2,700 s
Muse Wonder Full 64.6% 1,206 s
Claude Sonnet 5.5 64.5% 2,611 s
Claude Fable 5.2 63.6% 495.48 s
Qwen4-Max 62% 1,303 s
Qwen3.8 Max (0902) 61.7% 3,915 s
Inkling 2 Small 61.5% 514.76 s
GLM-5.3 61.4% 2,144 s
Muse Wonder 61.2% 612.47 s
Muse Ember 60.9% 156.44 s
DeepSeek V4.1 Flash 1017 60.9% 1,382 s
Gemini 3.8 Flash 60.5% 463.81 s
Grok 4.7 58.8% 1,249 s
Tencent Hy4 58.1% 2,369 s
Kimi K3 58% 1,344 s
MiMo-V2.6 Pro 57.9% 1,848 s
Tencent Hy4 preview 57.5% 1,935 s
Claude Haiku 5.5 57.5% 2,011 s
GPT-6.2 Astra 57.3% 2,412 s
MiMo-V2.6 Flash 57.2% 1,166 s
DeepSeek V4.1 Flash 57.1% 696.59 s
Mistral Large 4 57% 932.62 s
Mistral Large 4 Preview 56.9% 3,110 s
GPT-6.2 Sol 56.7% 1,607 s
Kimi K3.1 Pro 56% 2,300 s
MiniMax-M3.1-Flash 55.9% 1,572 s
GLM-5.3-Flash 55.8% 2,029 s
GPT-6.1 Sol 55.8% 1,156 s
GPT-6 Astra 55.6% 977.87 s
Step 5 Preview 55.4% 1,482 s
GPT-6.1 Terra 53.3% 1,765 s
GPT-6 Luna 53.1% 1,321 s
Muse Spark Open 53% 2,464 s
DeepSeek V4 Pro 0813 52.9% 1,440 s
MiniMax-M3 48.4% 600.9 s
Mercury 3 47.2% 344.44 s
Beam 1.0 46.5% 449.23 s
Nemotron 3.5 Super VL 41% 322.96 s
Inkling 38.3% 988.71 s
Inkling-Small 36.3% 624.36 s

The best score, 68.3%, costs $4.88 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta Thinking Machines Google DeepMind Qwen Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Neon Muse Wonder Full Inkling 2 Small Muse Ember 0% 20% 40% 60% 80% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Neon Qwen4-Max

Source: Official Vals Finance Index leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 4 Neon 68.3% $4.88 178,052
Gemini 4 Carbon 67% $6.46 160,960
Gemini 4 Hydrogen 66.1% $3.80 256,546
Gemini 4 Argon 66% $6.53 95,106
Claude Opus 5.5 65.3% $11 312,134
Muse Wonder Full 64.5% $3.14 57,946
Claude Sonnet 5.5 64.5% $7.40 395,199
Claude Fable 5.2 63.6% $4.94 48,712
Qwen4-Max 62% $1.26 98,481
Qwen3.8 Max (0902) 61.7% $3.83 197,086
Inkling 2 Small 61.5% $0.28 87,270
GLM-5.3 61.4% $2.21 169,379
Muse Wonder 61.2% $3.37 53,074
Muse Ember 60.9% $0.06 44,388
DeepSeek V4.1 Flash 1017 60.9% $0.42 215,664
Gemini 3.8 Flash 60.5% $3.70 102,660
Grok 4.7 58.8% $3.51 86,233
Tencent Hy4 58% $1.20 251,853
Kimi K3 58% $2.58 69,391
MiMo-V2.6 Pro 57.9% $0.22 87,303
Tencent Hy4 preview 57.5% $0.68 121,099
Claude Haiku 5.5 57.5% $1.39 482,804
GPT-6.2 Astra 57.3% $6.23 70,392
MiMo-V2.6 Flash 57.1% $0.07 84,370
DeepSeek V4.1 Flash 57.1% $0.25 132,306
Mistral Large 4 57% $0.56 67,899
Mistral Large 4 Preview 56.9% $1.46 125,008
GPT-6.2 Sol 56.7% $1.77 73,120
Kimi K3.1 Pro 56% $2.51 117,528
MiniMax-M3.1-Flash 55.9% $0.25 208,140
GLM-5.3-Flash 55.8% $0.55 159,905
GPT-6.1 Sol 55.8% $1.36 58,087
GPT-6 Astra 55.6% $6.48 55,355
Step 5 Preview 55.4% $0.82 91,378
GPT-6.1 Terra 53.3% $1.55 150,291
GPT-6 Luna 53% $0.15 141,399
Muse Spark Open 53% $3.00 44,557
DeepSeek V4 Pro 0813 52.9% $0.88 106,288
MiniMax-M3 48.4% $1.22 63,084
Mercury 3 47.2% $0.10 78,420
Beam 1.0 46.5% 39,315
Nemotron 3.5 Super VL 41% $0.48 93,952
Inkling 38.3% $0.97 63,992
Inkling-Small 36.3% $0.18 51,276

The best score, 68.3%, costs $4.88 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta Thinking Machines Google DeepMind Qwen Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Neon Muse Wonder Full Inkling 2 Small Muse Ember 0% 20% 40% 60% 80% $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Neon Qwen4-Max

Source: Official Vals Finance Index leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 4 Neon 68.3% $4.88 178,052
Gemini 4 Carbon 67% $6.46 160,960
Gemini 4 Hydrogen 66.1% $3.80 256,546
Gemini 4 Argon 66% $6.53 95,106
Claude Opus 5.5 65.3% $11 312,134
Muse Wonder Full 64.5% $3.14 57,946
Claude Sonnet 5.5 64.5% $7.40 395,199
Claude Fable 5.2 63.6% $4.94 48,712
Qwen4-Max 62% $1.26 98,481
Qwen3.8 Max (0902) 61.7% $3.83 197,086
Inkling 2 Small 61.5% $0.28 87,270
GLM-5.3 61.4% $2.21 169,379
Muse Wonder 61.2% $3.37 53,074
Muse Ember 60.9% $0.06 44,388
DeepSeek V4.1 Flash 1017 60.9% $0.42 215,664
Gemini 3.8 Flash 60.5% $3.70 102,660
Grok 4.7 58.8% $3.51 86,233
Tencent Hy4 58% $1.20 251,853
Kimi K3 58% $2.58 69,391
MiMo-V2.6 Pro 57.9% $0.22 87,303
Tencent Hy4 preview 57.5% $0.68 121,099
Claude Haiku 5.5 57.5% $1.39 482,804
GPT-6.2 Astra 57.3% $6.23 70,392
MiMo-V2.6 Flash 57.1% $0.07 84,370
DeepSeek V4.1 Flash 57.1% $0.25 132,306
Mistral Large 4 57% $0.56 67,899
Mistral Large 4 Preview 56.9% $1.46 125,008
GPT-6.2 Sol 56.7% $1.77 73,120
Kimi K3.1 Pro 56% $2.51 117,528
MiniMax-M3.1-Flash 55.9% $0.25 208,140
GLM-5.3-Flash 55.8% $0.55 159,905
GPT-6.1 Sol 55.8% $1.36 58,087
GPT-6 Astra 55.6% $6.48 55,355
Step 5 Preview 55.4% $0.82 91,378
GPT-6.1 Terra 53.3% $1.55 150,291
GPT-6 Luna 53% $0.15 141,399
Muse Spark Open 53% $3.00 44,557
DeepSeek V4 Pro 0813 52.9% $0.88 106,288
MiniMax-M3 48.4% $1.22 63,084
Mercury 3 47.2% $0.10 78,420
Beam 1.0 46.5% 39,315
Nemotron 3.5 Super VL 41% $0.48 93,952
Inkling 38.3% $0.97 63,992
Inkling-Small 36.3% $0.18 51,276

3 models have held number one on the official leaderboard since its release in October 2026 3 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. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.

Google DeepMind Thinking Machines
Own leaderboard 3 leaders
Lab claims 1 leader

Sources: Official Vals Finance Index leaderboard and labs’ model cards

Scores over time

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

Plot with other benchmarks
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%) Gemini 4 Neon Qwen3.8 Max (0902) MiniMax-M3 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Gemini 4 Neon Qwen3.8 Max (0902)

Sources: Official Vals Finance Index 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 20, 2026 Inkling 2 Mini 50.1 Vals Finance Index
Nov 5, 2026 Inkling 2 Small 61.5 Vals Finance Index

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Nov 20, 2026 Inkling 2 Mini model card 50.1 Vals Finance Index Capabilities

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 61.5% Vals Finance Index 61.5% Leaderboard Same

More in Economic work

All 210 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
  • Agents' Last Exam v1.1 In use Agents complete 122 public long-horizon professional workflow tasks across 55 sub-industries. Deterministic graders check outcomes against hidden references. GPT-6.2 Astra 47.5%
  • 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%

Data audit

All 12 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (45 entries), launch posts (2 scores) and system cards (1 score).
  • Pass Models scored 45 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
  • Pass Frontier models 9 of the top 10 have a score, led by GPT-6.2 Astra (#1).
  • Pass Headroom The best is 68.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on vals.ai: 44 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 19th, 2026.
  • Pass Read worked The last read worked.
Consistency 2 pass
  • Pass Lab claims The one pair agrees within 10 points.
  • Pass Known issues No published audit found problems.

How we check