Benchmark Almanac

CorpFin v2

Claude Opus 5 73.2% leads since Jul 2026 In use

Models answer expert-written questions about credit agreements that often exceed 200 pages. An LLM judge grades each answer for accuracy.

Vals archived it as saturated and stopped running new models (last update 2026-08-12). Splits are 20 public, 340 licensable validation and 858 private test questions from 61 credit agreements.

Made by Vals AI Finance Public tasks plus a private test set Last in a launch post Jul 2026 Last model added Aug 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Open-ended Free-form text, documents, designs or media judged as a whole.
Leaderboard vals.ai
119
Models scored on its own leaderboard
1
Launch post lists it, from 1 lab
1
System card reports it, from 1 lab

Claude Opus 5 leads with 73.2% among models released by Aug 2026

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

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

119 models. Scroll sideways for the rest.

  1. 73.2 Claude Opus 5
  2. 71.8 Claude Fable 5
  3. 71.6 Kimi K3
  4. 71.3 Muse Spark 1.1
  5. 70.9 Muse Spark 1.2
  6. 69.6 Inkling-Small
  7. 68.6 Inkling
  8. 68.5 Grok 4.3
  9. 68.4 GPT-5.5
  10. 68.3 Kimi K2.5
  11. 68.1 MiniMax-M3
  12. 68.0 Qwen3-Max
  13. 67.4 Grok 4.5
  14. 67.0 Claude Opus 4.6
  15. 67.0 Claude Sonnet 5
  16. 66.9 Grok 4 Fast
  17. 66.7 Kimi K2.6
  18. 66.7 Claude Opus 4.8
  19. 66.5 Qwen 3.6 Max (Preview)
  20. 66.4 Gemini 3 Flash
  21. 66.2 Grok 4.6
  22. 66.1 GLM-5.2
  23. 66.1 Claude Opus 4.7
  24. 66.0 Grok 4
  25. 66.0 Grok 4.1 Fast
  26. 65.9 GPT-5.2
  27. 65.8 Qwen 3.8 Max
  28. 65.5 Nemotron 3 Ultra
  29. 65.4 DeepSeek V4 Pro 0813
  30. 65.3 Qwen 3.5 Plus (hosted 397B-A17B)
  31. 65.3 Claude Sonnet 4.6
  32. 65.3 GPT-5.6 Terra
  33. 65.3 GPT-5.4
  34. 65.1 Muse Spark
  35. 65.1 Claude Opus 4.5
  36. 64.7 Gemini 3.5 Flash
  37. 64.5 Gemini 3.1 Pro
  38. 64.5 GLM-5.1
  39. 64.4 GPT-5.6 Sol
  40. 64.2 GPT-5.6 Luna
  41. 63.8 GPT-5.1
  42. 63.7 Qwen3.7-Max
  43. 63.7 Gemini 3 Pro
  44. 63.7 Grok 4.20
  45. 63.6 Qwen 3.5 Flash (hosted 35B-A3B)
  46. 63.3 Gemini 3.6 Flash
  47. 63.0 GPT-4.1
  48. 62.9 GLM-5
  49. 62.3 Qwen3.6 27B
  50. 62.0 Claude Sonnet 4.5
  51. 61.9 Qwen 3.6 Plus
  52. 61.9 Ling-3.0-flash
  53. 61.9 DeepSeek V4 Flash 0731
  54. 61.4 MiMo-V2.5-Pro
  55. 61.4 DeepSeek-V4-Pro
  56. 61.2 Claude Sonnet 4
  57. 61.2 GPT-5.4 Nano
  58. 61.2 MiniMax-M2.7
  59. 61.1 Grok-3 mini
  60. 61.1 GPT-5
  61. 61.0 Mistral Large 3
  62. 61.0 GLM-4.5
  63. 60.9 GPT-5.4 Mini
  64. 60.8 Gemini 2.5 Pro
  65. 60.7 Gemini 3.5 Flash-Lite
  66. 60.6 Claude Haiku 4.5
  67. 60.6 Kimi K2 Thinking
  68. 60.4 Claude 3.7 Sonnet
  69. 60.2 GPT-5 mini
  70. 59.9 MiMo-V2.5
  71. 59.8 Gemini 2.5 Flash
  72. 59.7 Grok 3
  73. 59.7 o3
  74. 59.6 MiniMax-M2.5
  75. 59.4 Gemini 3.1 Flash-Lite
  76. 59.0 o4-mini
  77. 58.9 MiniMax-M2.1
  78. 58.8 Mistral Medium 3.5
  79. 58.2 gpt-oss-120b
  80. 58.2 Laguna M.1
  81. 57.9 GPT-4.1 mini
  82. 57.6 Gemini 2.5 Flash-Lite
  83. 56.8 GLM-4.6
  84. 56.3 Laguna XS.2
  85. 54.7 DeepSeek-V3
  86. 54.7 Trinity Large Thinking
  87. 54.2 Nemotron 3.5 Lightning 30B-A3B
  88. 54.2 Gemini 2.5 Flash
  89. 54.1 DeepSeek-R1
  90. 53.6 Claude 3.5 Sonnet
  91. 53.1 gpt-oss-20b
  92. 51.5 DeepSeek-V3.1
  93. 51.1 Grok-2
  94. 51.0 DeepSeek-V3.2
  95. 50.8 Claude 3.5 Haiku
  96. 50.7 Mistral Medium 3.1
  97. 50.4 Kimi K2
  98. 49.7 Llama 4 Maverick
  99. 47.4 Magistral Medium 1.2
  100. 46.8 Llama 4 Scout
  101. 46.4 GLM-4.7
  102. 46.0 Cohere Command A
  103. 45.9 GPT-4o
  104. 45.5 GPT-4o mini
  105. 45.3 o3-mini
  106. 44.2 Mistral Small 3.1
  107. 44.0 Magistral Small 1.2
  108. 43.4 Gemini 2.0 Pro
  109. 42.1 GPT-4.1 nano
  110. 41.5 Jamba 1.6 Large
  111. 40.5 Gemini 1.5 Pro
  112. 39.4 Jamba 1.5 Large
  113. 38.9 Llama 3.1-70B
  114. 38.2 Gemini 1.5 Flash
  115. 38.0 Jamba 1.6 Mini
  116. 37.8 Llama 3.1-8B
  117. 33.9 Jamba 1.5 Mini
  118. 33.7 Gemini 2.0 Flash
  119. 28.6 Gemini 1.5 Flash
  1. 1 Claude Opus 5 73.2%
  2. 2 Claude Fable 5 71.8%
  3. 3 Kimi K3 71.6%
  4. 4 Muse Spark 1.1 71.3%
  5. 5 Muse Spark 1.2 70.9%
  6. 6 Inkling-Small 69.6%
  7. 7 Inkling 68.6%
  8. 8 Grok 4.3 68.5%
  9. 9 GPT-5.5 68.4%
  10. 10 Kimi K2.5 thinking 68.3%
  11. 11 MiniMax-M3 68.1%
  12. 12 Qwen3-Max 68%
  13. 13 Grok 4.5 67.4%
  14. 14 Claude Opus 4.6 thinking 67%
  15. 15 Claude Sonnet 5 67%
  16. 16 Grok 4 Fast reasoning 66.9%
  17. 17 Kimi K2.6 66.7%
  18. 18 Claude Opus 4.8 66.7%
  19. 19 Qwen 3.6 Max (Preview) 66.5%
  20. 20 Gemini 3 Flash 66.4%
  21. 21 Grok 4.6 66.2%
  22. 22 GLM-5.2 66.1%
  23. 23 Claude Opus 4.7 66.1%
  24. 24 Grok 4 66%
  25. 25 Grok 4.1 Fast reasoning 66%
  26. 26 GPT-5.2 65.9%
  27. 27 Qwen 3.8 Max 65.8%
  28. 28 Nemotron 3 Ultra 65.5%
  29. 29 DeepSeek V4 Pro 0813 65.4%
  30. 30 Qwen 3.5 Plus (hosted 397B-A17B) thinking 65.3%
  31. 31 Claude Sonnet 4.6 65.3%
  32. 32 GPT-5.6 Terra 65.3%
  33. 33 GPT-5.4 xhigh 65.3%
  34. 34 Muse Spark 65.1%
  35. 35 Claude Opus 4.5 thinking 65.1%
  36. 36 Gemini 3.5 Flash 64.7%
  37. 37 Gemini 3.1 Pro 64.5%
  38. 38 GLM-5.1 64.5%
  39. 39 GPT-5.6 Sol 64.4%
  40. 40 GPT-5.6 Luna 64.2%
  41. 41 GPT-5.1 63.8%
  42. 42 Qwen3.7-Max 63.7%
  43. 43 Gemini 3 Pro 63.7%
  44. 44 Grok 4.20 reasoning 63.7%
  45. 45 Qwen 3.5 Flash (hosted 35B-A3B) 63.6%
  46. 46 Gemini 3.6 Flash 63.3%
  47. 47 GPT-4.1 63%
  48. 48 GLM-5 thinking 62.9%
  49. 49 Qwen3.6 27B 62.3%
  50. 50 Claude Sonnet 4.5 thinking 62%
  51. 51 Qwen 3.6 Plus 61.9%
  52. 52 Ling-3.0-flash 61.9%
  53. 53 DeepSeek V4 Flash 0731 61.9%
  54. 54 MiMo-V2.5-Pro 61.4%
  55. 55 DeepSeek-V4-Pro 61.4%
  56. 56 Claude Sonnet 4 thinking 61.2%
  57. 57 GPT-5.4 Nano 61.2%
  58. 58 MiniMax-M2.7 61.2%
  59. 59 Grok-3 mini reasoning 61.1%
  60. 60 GPT-5 61.1%
  61. 61 Mistral Large 3 61%
  62. 62 GLM-4.5 61%
  63. 63 GPT-5.4 Mini 60.9%
  64. 64 Gemini 2.5 Pro Jun 2025 60.8%
  65. 65 Gemini 3.5 Flash-Lite 60.7%
  66. 66 Claude Haiku 4.5 thinking 60.6%
  67. 67 Kimi K2 Thinking 60.6%
  68. 68 Claude 3.7 Sonnet thinking 60.4%
  69. 69 GPT-5 mini 60.2%
  70. 70 MiMo-V2.5 59.9%
  71. 71 Gemini 2.5 Flash Sep 2025, thinking 59.8%
  72. 72 Grok 3 59.7%
  73. 73 o3 59.7%
  74. 74 MiniMax-M2.5 59.6%
  75. 75 Gemini 3.1 Flash-Lite 59.4%
  76. 76 o4-mini 59%
  77. 77 MiniMax-M2.1 58.9%
  78. 78 Mistral Medium 3.5 58.8%
  79. 79 gpt-oss-120b 58.2%
  80. 80 Laguna M.1 58.2%
  81. 81 GPT-4.1 mini 57.9%
  82. 82 Gemini 2.5 Flash-Lite Sep 2025, thinking 57.6%
  83. 83 GLM-4.6 56.8%
  84. 84 Laguna XS.2 56.3%
  85. 85 DeepSeek-V3 Mar 2025 54.7%
  86. 86 Trinity Large Thinking 54.7%
  87. 87 Nemotron 3.5 Lightning 30B-A3B 54.2%
  88. 88 Gemini 2.5 Flash Apr 2025, non-thinking 54.2%
  89. 89 DeepSeek-R1 May 2025 54.1%
  90. 90 Claude 3.5 Sonnet 53.6%
  91. 91 gpt-oss-20b 53.1%
  92. 92 DeepSeek-V3.1 51.5%
  93. 93 Grok-2 Dec 2024 51.1%
  94. 94 DeepSeek-V3.2 thinking 51%
  95. 95 Claude 3.5 Haiku 50.8%
  96. 96 Mistral Medium 3.1 50.7%
  97. 97 Kimi K2 Jul 2025 50.4%
  98. 98 Llama 4 Maverick 49.7%
  99. 99 Magistral Medium 1.2 47.4%
  100. 100 Llama 4 Scout 46.8%
  101. 101 GLM-4.7 46.4%
  102. 102 Cohere Command A 46%
  103. 103 GPT-4o Aug 2024 45.9%
  104. 104 GPT-4o mini 45.5%
  105. 105 o3-mini 45.3%
  106. 106 Mistral Small 3.1 44.2%
  107. 107 Magistral Small 1.2 44%
  108. 108 Gemini 2.0 Pro 43.4%
  109. 109 GPT-4.1 nano 42.1%
  110. 110 Jamba 1.6 Large 41.5%
  111. 111 Gemini 1.5 Pro Sep 2024 40.5%
  112. 112 Jamba 1.5 Large 39.4%
  113. 113 Llama 3.1-70B 38.9%
  114. 114 Gemini 1.5 Flash Sep 2024 38.2%
  115. 115 Jamba 1.6 Mini 38%
  116. 116 Llama 3.1-8B 37.8%
  117. 117 Jamba 1.5 Mini 33.9%
  118. 118 Gemini 2.0 Flash Feb 2025 33.7%
  119. 119 Gemini 1.5 Flash May 2024 28.6%
#ModelScoreSettingDateSource
1 Claude Opus 5 73.2% Jul 2026 Official leaderboard
2 Claude Fable 5 71.8% Jun 2026 Official leaderboard
3 Kimi K3 71.6% Jul 2026 Official leaderboard
4 Muse Spark 1.1 71.3% Jul 2026 Official leaderboard
5 Muse Spark 1.2 70.9% Aug 2026 Official leaderboard
6 Inkling-Small 69.6% Jul 2026 Official leaderboard
7 Inkling 68.6% Jul 2026 Official leaderboard
8 Grok 4.3 68.5% Apr 2026 Official leaderboard
9 GPT-5.5 68.4% Apr 2026 Official leaderboard
10 Kimi K2.5 68.3% thinking Jan 2026 Official leaderboard
11 MiniMax-M3 68.1% Jun 2026 Official leaderboard
12 Qwen3-Max 68% Sep 2025 Official leaderboard
13 Grok 4.5 67.4% Jul 2026 Official leaderboard
14 Claude Opus 4.6 67% thinking Feb 2026 Official leaderboard
15 Claude Sonnet 5 67% Jun 2026 Official leaderboard
16 Grok 4 Fast 66.9% reasoning Sep 2025 Official leaderboard
17 Kimi K2.6 66.7% Apr 2026 Official leaderboard
18 Claude Opus 4.8 66.7% May 2026 Official leaderboard
19 Qwen 3.6 Max (Preview) 66.5% Apr 2026 Official leaderboard
20 Gemini 3 Flash 66.4% Dec 2025 Official leaderboard
21 Grok 4.6 66.2% Aug 2026 Official leaderboard
22 GLM-5.2 66.1% Jun 2026 Official leaderboard
23 Claude Opus 4.7 66.1% Apr 2026 Official leaderboard
24 Grok 4 66% Jul 2025 Official leaderboard
25 Grok 4.1 Fast 66% reasoning Nov 2025 Official leaderboard
26 GPT-5.2 65.9% Dec 2025 Official leaderboard
27 Qwen 3.8 Max 65.8% Aug 2026 Official leaderboard
28 Nemotron 3 Ultra 65.5% Jun 2026 Official leaderboard
29 DeepSeek V4 Pro 0813 65.4% Aug 2026 Official leaderboard
30 Qwen 3.5 Plus (hosted 397B-A17B) 65.3% thinking Feb 2026 Official leaderboard
31 Claude Sonnet 4.6 65.3% Feb 2026 Official leaderboard
32 GPT-5.6 Terra 65.3% Jul 2026 Official leaderboard
33 GPT-5.4 65.3% xhigh Mar 2026 Official leaderboard
34 Muse Spark 65.1% Apr 2026 Official leaderboard
35 Claude Opus 4.5 65.1% thinking Nov 2025 Official leaderboard
36 Gemini 3.5 Flash 64.7% May 2026 Official leaderboard
37 Gemini 3.1 Pro 64.5% Feb 2026 Official leaderboard
38 GLM-5.1 64.5% Apr 2026 Official leaderboard
39 GPT-5.6 Sol 64.4% Jul 2026 Official leaderboard
40 GPT-5.6 Luna 64.2% Jul 2026 Official leaderboard
41 GPT-5.1 63.8% Nov 2025 Official leaderboard
42 Qwen3.7-Max 63.7% May 2026 Official leaderboard
43 Gemini 3 Pro 63.7% Nov 2025 Official leaderboard
44 Grok 4.20 63.7% reasoning Feb 2026 Official leaderboard
45 Qwen 3.5 Flash (hosted 35B-A3B) 63.6% Feb 2026 Official leaderboard
46 Gemini 3.6 Flash 63.3% Jul 2026 Official leaderboard
47 GPT-4.1 63% Apr 2025 Official leaderboard
48 GLM-5 62.9% thinking Feb 2026 Official leaderboard
49 Qwen3.6 27B 62.3% Apr 2026 Official leaderboard
50 Claude Sonnet 4.5 62% thinking Sep 2025 Official leaderboard
51 Qwen 3.6 Plus 61.9% Mar 2026 Official leaderboard
52 Ling-3.0-flash 61.9% Aug 2026 Official leaderboard
53 DeepSeek V4 Flash 0731 61.9% Jul 2026 Official leaderboard
54 MiMo-V2.5-Pro 61.4% Apr 2026 Official leaderboard
55 DeepSeek-V4-Pro 61.4% Apr 2026 Official leaderboard
56 Claude Sonnet 4 61.2% thinking May 2025 Official leaderboard
57 GPT-5.4 Nano 61.2% Mar 2026 Official leaderboard
58 MiniMax-M2.7 61.2% Mar 2026 Official leaderboard
59 Grok-3 mini 61.1% reasoning Apr 2025 Official leaderboard
60 GPT-5 61.1% Aug 2025 Official leaderboard
61 Mistral Large 3 61% Dec 2025 Official leaderboard
62 GLM-4.5 61% Aug 2025 Official leaderboard
63 GPT-5.4 Mini 60.9% Mar 2026 Official leaderboard
64 Gemini 2.5 Pro Jun 2025 60.8% Jun 2025 Official leaderboard
65 Gemini 3.5 Flash-Lite 60.7% Jul 2026 Official leaderboard
66 Claude Haiku 4.5 60.6% thinking Oct 2025 Official leaderboard
67 Kimi K2 Thinking 60.6% Nov 2025 Official leaderboard
68 Claude 3.7 Sonnet 60.4% thinking Feb 2025 Official leaderboard
69 GPT-5 mini 60.2% Aug 2025 Official leaderboard
70 MiMo-V2.5 59.9% Apr 2026 Official leaderboard
71 Gemini 2.5 Flash Sep 2025 59.8% thinking Sep 2025 Official leaderboard
72 Grok 3 59.7% Apr 2025 Official leaderboard
73 o3 59.7% Apr 2025 Official leaderboard
74 MiniMax-M2.5 59.6% Feb 2026 Official leaderboard
75 Gemini 3.1 Flash-Lite 59.4% Mar 2026 Official leaderboard
76 o4-mini 59% Apr 2025 Official leaderboard
77 MiniMax-M2.1 58.9% Dec 2025 Official leaderboard
78 Mistral Medium 3.5 58.8% Apr 2026 Official leaderboard
79 gpt-oss-120b 58.2% Aug 2025 Official leaderboard
80 Laguna M.1 58.2% Apr 2026 Official leaderboard
81 GPT-4.1 mini 57.9% Apr 2025 Official leaderboard
82 Gemini 2.5 Flash-Lite Sep 2025 57.6% thinking Sep 2025 Official leaderboard
83 GLM-4.6 56.8% Sep 2025 Official leaderboard
84 Laguna XS.2 56.3% Apr 2026 Official leaderboard
85 DeepSeek-V3 Mar 2025 54.7% Mar 2025 Official leaderboard
86 Trinity Large Thinking 54.7% Apr 2026 Official leaderboard
87 Nemotron 3.5 Lightning 30B-A3B 54.2% Aug 2026 Official leaderboard
88 Gemini 2.5 Flash Apr 2025 54.2% non-thinking Apr 2025 Official leaderboard
89 DeepSeek-R1 May 2025 54.1% May 2025 Official leaderboard
90 Claude 3.5 Sonnet 53.6% Jun 2024 Official leaderboard
91 gpt-oss-20b 53.1% Aug 2025 Official leaderboard
92 DeepSeek-V3.1 51.5% Aug 2025 Official leaderboard
93 Grok-2 Dec 2024 51.1% Dec 2024 Official leaderboard
94 DeepSeek-V3.2 51% thinking Dec 2025 Official leaderboard
95 Claude 3.5 Haiku 50.8% Oct 2024 Official leaderboard
96 Mistral Medium 3.1 50.7% Aug 2025 Official leaderboard
97 Kimi K2 Jul 2025 50.4% Jul 2025 Official leaderboard
98 Llama 4 Maverick 49.7% Apr 2025 Official leaderboard
99 Magistral Medium 1.2 47.4% Sep 2025 Official leaderboard
100 Llama 4 Scout 46.8% Apr 2025 Official leaderboard
101 GLM-4.7 46.4% Dec 2025 Official leaderboard
102 Cohere Command A 46% Mar 2025 Official leaderboard
103 GPT-4o Aug 2024 45.9% Aug 2024 Official leaderboard
104 GPT-4o mini 45.5% Jul 2024 Official leaderboard
105 o3-mini 45.3% Jan 2025 Official leaderboard
106 Mistral Small 3.1 44.2% Mar 2025 Official leaderboard
107 Magistral Small 1.2 44% Sep 2025 Official leaderboard
108 Gemini 2.0 Pro 43.4% Feb 2025 Official leaderboard
109 GPT-4.1 nano 42.1% Apr 2025 Official leaderboard
110 Jamba 1.6 Large 41.5% Mar 2025 Official leaderboard
111 Gemini 1.5 Pro Sep 2024 40.5% Sep 2024 Official leaderboard
112 Jamba 1.5 Large 39.4% Aug 2024 Official leaderboard
113 Llama 3.1-70B 38.9% Jul 2024 Official leaderboard
114 Gemini 1.5 Flash Sep 2024 38.2% Aug 2024 Official leaderboard
115 Jamba 1.6 Mini 38% Mar 2025 Official leaderboard
116 Llama 3.1-8B 37.8% Jul 2024 Official leaderboard
117 Jamba 1.5 Mini 33.9% Aug 2024 Official leaderboard
118 Gemini 2.0 Flash Feb 2025 33.7% Feb 2025 Official leaderboard
119 Gemini 1.5 Flash May 2024 28.6% May 2024 Official leaderboard

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

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Grok 4 Fast
non-reasoni… reasoning
66.9%
Grok 4.1 Fast
non-reasoni… reasoning
66%
Claude Opus 4.5
non-thinking thinking
65.1%
Claude Sonnet 4.5
non-thinking
62%
Claude Sonnet 4
non-thinking thinking
61.2%
Claude Haiku 4.5
non-thinking
60.6%
Gemini 2.5 Flash
non-thinking
59.8%
Gemini 2.5 Flash-Lite
non-thinking
57.6%
DeepSeek-V3.2
non-thinking thinking
51%
ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5 73.2%
Claude Fable 5 71.8%
Kimi K3 71.6%
Muse Spark 1.1 71.3%
Muse Spark 1.2 70.9%
Inkling-Small 69.6%
Inkling 68.6%
Grok 4.3 68.5%
GPT-5.5 68.4%
Kimi K2.5 thinking 68.3%
MiniMax-M3 68.1%
Qwen3-Max 68%
Grok 4.5 67.4%
Claude Opus 4.6 thinking 67%
Claude Sonnet 5 67%
Grok 4 Fast reasoning 66.9%
Kimi K2.6 66.7%
Claude Opus 4.8 66.7%
Qwen 3.6 Max (Preview) 66.5%
Gemini 3 Flash 66.4%
Grok 4.6 66.2%
GLM-5.2 66.1%
Claude Opus 4.7 66.1%
Grok 4 66%
Grok 4.1 Fast reasoning 66%
GPT-5.2 65.9%
Qwen 3.8 Max 65.8%
Nemotron 3 Ultra 65.5%
DeepSeek V4 Pro 0813 65.4%
Qwen 3.5 Plus (hosted 397B-A17B) thinking 65.3%
Claude Sonnet 4.6 65.3%
GPT-5.6 Terra 65.3%
GPT-5.4 xhigh 65.3%
Muse Spark 65.1%
Claude Opus 4.5 thinking 65.1%
Gemini 3.5 Flash 64.7%
Gemini 3.1 Pro 64.5%
GLM-5.1 64.5%
GPT-5.6 Sol 64.4%
GPT-5.6 Luna 64.2%
GPT-5.1 63.8%
Qwen3.7-Max 63.7%
Gemini 3 Pro 63.7%
Grok 4.20 reasoning 63.7%
Qwen 3.5 Flash (hosted 35B-A3B) 63.6%
Gemini 3.6 Flash 63.3%
GPT-4.1 63%
GLM-5 thinking 62.9%
Qwen3.6 27B 62.3%
Claude Sonnet 4.5 thinking 62%
Qwen 3.6 Plus 61.9%
Ling-3.0-flash 61.9%
DeepSeek V4 Flash 0731 61.9%
MiMo-V2.5-Pro 61.4%
DeepSeek-V4-Pro 61.4%
Claude Opus 4.5 non-thinking 61.3%
Claude Sonnet 4 thinking 61.2%
GPT-5.4 Nano 61.2%
MiniMax-M2.7 61.2%
Grok-3 mini reasoning 61.1%
GPT-5 61.1%
Mistral Large 3 61%
GLM-4.5 61%
GPT-5.4 Mini 60.9%
Gemini 2.5 Pro Jun 2025 60.8%
Claude Sonnet 4.5 non-thinking 60.8%
Gemini 3.5 Flash-Lite 60.7%
Claude Haiku 4.5 thinking 60.6%
Kimi K2 Thinking 60.6%
Claude 3.7 Sonnet thinking 60.4%
Claude Haiku 4.5 non-thinking 60.3%
GPT-5 mini 60.2%
MiMo-V2.5 59.9%
Gemini 2.5 Pro Jun 2025 59.8%
Gemini 2.5 Flash Sep 2025 thinking 59.8%
Grok 3 59.7%
o3 59.7%
MiniMax-M2.5 59.6%
Grok-3 mini reasoning 59.5%
Gemini 3.1 Flash-Lite 59.4%
o4-mini 59%
Gemini 2.5 Flash Sep 2025 non-thinking 59%
MiniMax-M2.1 58.9%
Mistral Medium 3.5 58.8%
Grok 4 Fast non-reasoning 58.4%
gpt-oss-120b 58.2%
Laguna M.1 58.2%
GPT-4.1 mini 57.9%
Gemini 2.5 Flash-Lite Sep 2025 thinking 57.6%
GLM-4.6 56.8%
Laguna XS.2 56.3%
Gemini 2.5 Flash-Lite Sep 2025 non-thinking 56.3%
Qwen3-Max 55.9%
DeepSeek-V3 Mar 2025 54.7%
Claude Sonnet 4 non-thinking 54.7%
Trinity Large Thinking 54.7%
Nemotron 3.5 Lightning 30B-A3B 54.2%
Gemini 2.5 Flash Apr 2025 non-thinking 54.2%
DeepSeek-R1 May 2025 54.1%
Claude 3.5 Sonnet 53.6%
gpt-oss-20b 53.1%
Qwen3-Max 53%
DeepSeek-V3 Mar 2025 52.5%
Grok 4.1 Fast non-reasoning 52.5%
DeepSeek-V3.1 51.5%
Grok-2 Dec 2024 51.1%
DeepSeek-V3.2 thinking 51%
Claude 3.5 Haiku 50.8%
Mistral Medium 3.1 50.7%
Kimi K2 Jul 2025 50.4%
Llama 4 Maverick 49.7%
DeepSeek-V3.2 non-thinking 47.9%
Magistral Medium 1.2 47.4%
Llama 4 Scout 46.8%
GLM-4.7 46.4%
Cohere Command A 46%
GPT-4o Aug 2024 45.9%
GPT-4o mini 45.5%
o3-mini 45.3%
Mistral Small 3.1 44.2%
Magistral Small 1.2 44%
Gemini 2.0 Pro 43.4%
GPT-4.1 nano 42.1%
Jamba 1.6 Large 41.5%
Gemini 1.5 Pro Sep 2024 40.5%
GPT-4o Aug 2024 39.4%
Jamba 1.5 Large 39.4%
Llama 3.1-70B 38.9%
Gemini 1.5 Flash Sep 2024 38.2%
Jamba 1.6 Mini 38%
Llama 3.1-8B 37.8%
Jamba 1.5 Mini 33.9%
Gemini 2.0 Flash Feb 2025 33.7%
Gemini 1.5 Flash May 2024 28.6%

13 models have held number one since 2023

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

Google DeepMind Anthropic xAI OpenAI Qwen Moonshot AI
Own leaderboard Not tested since Aug 2026
Claude Opus 5 73.2%

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% 2024 2025 2026 Near the ceiling (85%) Claude Opus 5 Grok 4.3 Qwen3-Max GPT-4.1 Claude 3.5 Sonnet Gemini 1.5 Flash 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Claude Opus 5 Grok 4 Fast

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.

Moonshot AI
DateLaunch postScoreAs the post writes it
16 Jul 2026 Kimi K3 71.6 CorpFin v2

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
27 Jul 2026 Kimi K3 technical report 71.6 CorpFin v2, cited from Vals AI (Table 2) Capabilities

More in Law, finance and medicine

All 140 in Law, finance and medicine
  • HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
  • MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy. Qwen 3.8 Max 80.4%
  • HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
  • KBL In use Models answer Korean legal knowledge and reasoning tasks and Korean bar exam questions. It has 7 knowledge tasks, 4 reasoning tasks and 2,510 bar exam items. Solar Pro 4 77.5%
  • KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. Solar Pro 4 93.2%
  • PRBench In use Model answers 1,100 expert-written law and finance questions. LLM judge grades against weighted expert rubrics of 10-30 criteria each. Muse Spark 1.3 61.6%