Skip to content

Benchmarks · Vals AI

CorpFin v2

Measures understanding of long credit agreements (question answering over contract text; private dataset). Vals marks the leaderboard "Archived Benchmark" (saturated; no new models; last update 12 August 2026).

Models measured
108
Top
73.19 %Claude Opus 5within the standard error of ±0.87
Version
2
Measured
12 August 2026
Go to the sourceThis benchmark does not count towards the PrudentScore.

01

Ranking

At the top: Top 15

Vals AI Legal Research Bench, all-pass [1]. Hatched bars are rejected by the compliance gate.

  1. 1Claude Opus 573.19%
  2. 2Claude Fable 571.83%
  3. 3Kimi K371.56%
  4. 4Muse Spark 1.171.29%
  5. 5Muse Spark 1.270.94%
  6. 6Inkling Small69.62%
  7. 7Inkling68.57%
  8. 8Grok 4.368.53%
  9. 9GPT 5.568.42%
  10. 10Kimi K2.568.26%
  11. 11MiniMax-M368.10%
  12. 12Qwen 3 Max Thinking68.03%
  13. 13Grok 4.567.41%
  14. 14Claude Opus 4.6 (Thinking)67.02%
  15. 15Claude Sonnet 566.98%

Source: Vals AI (CorpFin v2) · Licence: CC BY 4.0 · Adaptation: Prudai selected the models, translated the explanations, added the gate verdict and the PrudentScore and rounded figures to two decimals.

Passes the compliance gateRejected by the compliance gate
CorpFin v2 2
# ModelScoreCost/taskGatePrudentScore
1Claude Opus 5Anthropic73.19%± 0.87-Passes72
2Claude Fable 5Anthropic71.83%± 0.88-Rejectedno score
3Kimi K3Moonshot AI71.56%± 0.89-Rejectedno score
4Muse Spark 1.1Meta71.29%± 0.89-Rejectedno score
5Muse Spark 1.2Meta70.94%± 0.89-Rejectedno score
6Inkling SmallThinking Machines69.62%± 0.91-Rejectedno score
7InklingThinking Machines68.57%± 0.92-Rejectedno score
8Grok 4.3xAI (SpaceXAI)68.53%± 0.92-Rejectedno score
9GPT 5.5OpenAI68.42%± 0.92-Passes51
10Kimi K2.5Moonshot AI68.26%± 0.92-Passes37
11MiniMax-M3MiniMax68.10%± 0.92-Rejectedno score
12Qwen 3 Max ThinkingAlibaba68.03%± 0.92-Not yet assessedno score
13Grok 4.5xAI (SpaceXAI)67.41%± 0.92-Rejectedno score
14Claude Opus 4.6 (Thinking)Anthropic67.02%± 0.93-Passesno score
15Claude Sonnet 5Anthropic66.98%± 0.93-Passes59
16Grok 4 Fast (Reasoning)xAI (SpaceXAI)66.90%± 0.93-Rejectedno score
17Kimi K2.6Moonshot AI66.74%± 0.93-Rejectedno score
18Claude Opus 4.8Anthropic66.71%± 0.93-Passes62
19Qwen 3.6 Max PreviewAlibaba66.47%± 0.93-Not yet assessedno score
20Gemini 3 Flash (12/25)Google66.43%± 0.93-Rejectedno score
21Grok 4.6xAI (SpaceXAI)66.16%± 0.93-Rejectedno score
22GLM 5.2Z.ai66.12%± 0.93-Rejectedno score
23Claude Opus 4.7Anthropic66.08%± 0.93-Passes54
24Grok 4xAI (SpaceXAI)66.05%± 0.93-Rejectedno score
25Grok 4.1 Fast (Reasoning)xAI (SpaceXAI)65.97%± 0.93-Rejectedno score
26GPT 5.2OpenAI65.89%± 0.93-Passesno score
27Qwen 3.8 MaxAlibaba65.85%± 0.94-Not yet assessedno score
28Nemotron 3 UltraNVIDIA65.46%± 0.94-Rejectedno score
29DeepSeek V4 Pro (0813)DeepSeek65.42%± 0.94-Rejectedno score
30=Claude Sonnet 4.6Anthropic65.31%± 0.94-Passes55
30=GPT-5.6 TerraOpenAI65.31%± 0.94-Passes53
30=Qwen 3.5 PlusAlibaba65.31%± 0.94-Not yet assessedno score
33GPT 5.4 (xhigh)OpenAI65.27%± 0.94-Passes0
34Muse SparkMeta65.11%± 0.94-Rejectedno score
35Claude Opus 4.5 (Thinking)Anthropic65.07%± 0.94-Passesno score
36Gemini 3.5 FlashGoogle64.69%± 0.94-Passes47
37Gemini 3.1 Pro Preview (02/26)Google64.49%± 0.94-Rejectedno score
38GLM 5.1Z.ai64.45%± 0.94-Rejectedno score
39GPT-5.6 SolOpenAI64.38%± 0.94-Passes57
40GPT-5.6 LunaOpenAI64.22%± 0.94-Passes56
41GPT 5.1OpenAI63.83%± 0.95-Passesno score
42Qwen 3.7 MaxAlibaba63.71%± 0.95-Not yet assessedno score
43=Gemini 3 Pro (11/25)Google63.67%± 0.95-Rejectedno score
43=Grok 4.20 (Reasoning)xAI (SpaceXAI)63.67%± 0.95-Rejectedno score
45Qwen 3.5 FlashAlibaba63.56%± 0.95-Not yet assessedno score
46Gemini 3.6 FlashGoogle63.33%± 0.95-Passes46
47GPT 4.1OpenAI63.05%± 0.95-Passesno score
48GLM 5Z.ai62.90%± 0.95-Passesno score
49Qwen 3.6 27BAlibaba62.31%± 0.95-Rejectedno score
50Claude Sonnet 4.5 (Thinking)Anthropic61.97%± 0.96-Passesno score
51=Ling 3.0 FlashAnt Group61.93%± 0.96-Rejectedno score
51=Qwen 3.6 PlusAlibaba61.93%± 0.96-Not yet assessedno score
53DeepSeek V4 Flash 0731DeepSeek61.85%± 0.96-Rejectedno score
54MiMo V2.5 ProXiaomi61.42%± 0.96-Rejectedno score
55DeepSeek V4 Pro (2026-04-23)DeepSeek61.38%± 0.96-Rejectedno score
56Claude Opus 4.5 (Nonthinking)Anthropic61.30%± 0.96-Passesno score
57Claude Sonnet 4 (Thinking)Anthropic61.23%± 0.96-Rejectedno score
58=GPT 5.4 NanoOpenAI61.19%± 0.96-Passes29
58=MiniMax-M2.7MiniMax61.19%± 0.96-Rejectedno score
60GPT 5OpenAI61.07%± 0.96-Passesno score
61Mistral Large 3Mistral AI61.03%± 0.96-Passesno score
62GLM 4.5Z.ai60.96%± 0.96-Rejectedno score
63GPT 5.4 MiniOpenAI60.92%± 0.96-Passes25
64=Claude Sonnet 4.5 (Nonthinking)Anthropic60.80%± 0.96-Passesno score
64=Gemini 2.5 ProGoogle60.80%± 0.96-Not yet assessedno score
66Gemini 3.5 Flash LiteGoogle60.68%± 0.96-Passes38
67Claude Haiku 4.5 (Thinking)Anthropic60.61%± 0.96-Passes32
68Kimi K2 ThinkingMoonshot AI60.57%± 0.96-Rejectedno score
69Claude Haiku 4.5 (Nonthinking)Anthropic60.30%± 0.96-Passesno score
70GPT 5 MiniOpenAI60.18%± 0.96-Passesno score
71MiMo V2.5Xiaomi59.91%± 0.97-Rejectedno score
72Gemini 2.5 Flash Preview (9/25) (Thinking)Google59.75%± 0.97-Rejectedno score
73o3OpenAI59.71%± 0.96-Passesno score
74MiniMax-M2.5MiniMax59.60%± 0.97-Passesno score
75Gemini 3.1 Flash Lite PreviewGoogle59.36%± 0.97-Not yet assessedno score
76=Gemini 2.5 Flash Preview (9/25) (Nonthinking)Google58.97%± 0.97-Rejectedno score
76=o4 MiniOpenAI58.97%± 0.97-Passesno score
78MiniMax-M2.1MiniMax58.90%± 0.97-Passesno score
79Mistral Medium 3.5Mistral AI58.78%± 0.97-Passes23
80Grok 4 Fast (Non-Reasoning)xAI (SpaceXAI)58.39%± 0.97-Rejectedno score
81GPT OSS 120BOpenAI58.24%± 0.97-Passesno score
82Laguna M.1Poolside58.16%± 0.97-Rejectedno score
83GPT 4.1 MiniOpenAI57.93%± 0.97-Passesno score
84Gemini 2.5 Flash Lite (9/25) (Thinking)Google57.58%± 0.97-Rejectedno score
85GLM 4.6Z.ai56.84%± 0.97-Rejectedno score
86Laguna XS.2Poolside56.33%± 0.98-Rejectedno score
87Gemini 2.5 Flash Lite (9/25) (Nonthinking)Google56.29%± 0.98-Rejectedno score
88Qwen 3 MaxAlibaba55.94%± 0.98-Not yet assessedno score
89Claude Sonnet 4 (Nonthinking)Anthropic54.70%± 0.98-Rejectedno score
90Trinity Large ThinkingArcee AI54.66%± 0.98-Rejectedno score
91Nemotron 3.5 LightningNVIDIA54.23%± 0.98-Rejectedno score
92GPT OSS 20BOpenAI53.15%± 0.98-Passesno score
93Qwen 3 Max PreviewAlibaba52.95%± 0.97-Not yet assessedno score
94Grok 4.1 Fast Non-ReasoningxAI (SpaceXAI)52.49%± 0.97-Rejectedno score
95DeepSeek V3.1DeepSeek51.48%± 0.98-Passesno score
96DeepSeek V3.2 (Thinking)DeepSeek50.97%± 0.98-Passesno score
97Mistral Medium 3.1 (05/2025)Mistral AI50.74%± 0.97-Passesno score
98Kimi K2 InstructMoonshot AI50.39%± 0.98-Rejectedno score
99Llama 4 MaverickMeta49.73%± 0.98-Rejectedno score
100DeepSeek V3.2 (Nonthinking)DeepSeek47.94%± 0.98-Passesno score
101Magistral Medium 1.2 (09/2025)Mistral AI47.40%± 0.98-Not yet assessedno score
102Llama 4 ScoutMeta46.78%± 0.98-Rejectedno score
103GLM 4.7Z.ai46.39%± 0.90-Passesno score
104Command ACohere45.96%± 0.97-Rejectedno score
105GPT 4o (2024-11-20)OpenAI45.92%± 0.97-Passesno score
106Mistral Small 3.1 (03/2025)Mistral AI44.17%± 0.97-Passesno score
107Magistral Small 1.2 (09/2025)Mistral AI44.02%± 0.96-Passesno score
108GPT 4.1 NanoOpenAI42.08%± 0.97-Passesno score

Source: Vals AI (CorpFin v2) · Licence: CC BY 4.0 · Adaptation: Prudai selected the models, translated the explanations, added the gate verdict and the PrudentScore and rounded figures to two decimals.

Vals AI · Version 2 · Measured 12 August 2026 · Go to the source. The PrudentScore is a composition by Prudai B.V. based on publicly published figures and does not represent the view or endorsement of Vals AI, Harvey or Artificial Analysis. Figures from Artificial Analysis appear in running text only, with attribution and a link, as that party's terms require [9].

Other benchmarks

Claude and Anthropic, GPT and OpenAI, Gemini and Google, Grok and xAI, Kimi and Moonshot AI, Muse and Meta, Mistral, DeepSeek, GLM and Z.ai, Harvey, Vals AI and Artificial Analysis are trademarks of their respective owners. They are named for identification only and this implies no partnership, sponsorship or endorsement.