Add a benchmark

Benchmarks / Professional

CorpFin (Vals)

A private benchmark evaluating understanding of long-context credit agreements Scores are the values published by VALS AI. Source benchmark version 2.

115models
–Open Index weight
2026-08-12last updated
Vals AIpublisher

CorpFin (Vals)

Higher is better · best effort per model

Chart settings
115 of 115 selected

full table below

#ModelEffortScoreCost / taskHarness
1Opus 5max73.2%$0.860–
2Fable 5max71.8%$1.74–
3Kimi K3max71.6%$0.064–
4Muse Spark 1.1xhigh71.3%$0.109–
5Muse Spark 1.2xhigh70.9%$0.108–
6Inkling Smallxhigh69.6%$0.025–
7Inklingmax68.6%$0.084–
8Grok 4.3xhigh68.5%$0.134–
9GPT-5.5xhigh68.4%$0.432–
10Kimi K2.5max68.3%$0.052–
11MiniMax M3max68.1%$0.050–
12Qwen 3 Maxmax68.0%$0.261–
13Grok 4.5high67.4%$0.221–
14Opus 4.6max67.0%$0.487–
15Sonnet 5max67.0%$0.516–
16Grok 4 Fastnone66.9%$0.040–
17Kimi K2.6max66.7%$0.073–
18Opus 4.8max66.7%$0.856–
19Qwen 3.6 Maxmax66.5%$0.156–
20Gemini 3 Flashhigh66.4%$0.047–
21Grok 4.6high66.2%$0.166–
22GLM 5.2max66.1%$0.122–
23Opus 4.7max66.1%$0.816–
24Grok 4high66.0%$0.191–
25Grok 4.1 Fasthigh66.0%$0.016–
26GPT-5.2xhigh65.9%$0.154–
27Qwen 3.8 Maxmax65.9%$0.179–
28Nemotron 3 Ultra 550B A55Bmax65.5%––
29DeepSeek V4 Pro 0813max65.4%$0.0022–
30GPT-5.6 Terraxhigh65.3%$0.171–
31Sonnet 4.6max65.3%$0.307–
32Qwen 3.5 Plusnone65.3%$0.058–
33GPT-5.4xhigh65.3%$0.235–
34Muse Spark–65.1%$0.012–
35Opus 4.5high65.1%$0.173–
36Gemini 3.5 Flashhigh64.7%$0.140–
37Gemini 3.1 Prohigh64.5%$0.182–
38GLM 5.1max64.5%$0.079–
39GPT-5.6 Solmax64.4%$0.485–
40GPT-5.6 Lunamax64.2%$0.019–
41GPT-5.1high63.8%$0.092–
42Qwen 3.7 Maxmax63.7%$0.220–
43Grok 4.20xhigh63.7%$0.163–
44Gemini 3 Prohigh63.7%$0.179–
45Qwen 3.5 Flashmax63.6%$0.0096–
46Gemini 3.6 Flashhigh63.3%$0.137–
47GPT-4.1high63.1%$0.132–
48GLM 5high62.9%$0.068–
49Qwen 3.6 27Bmax62.3%$0.054–
50Sonnet 4.5high62.0%$0.255–
51Ling 3.0 Flash–61.9%$0.0064–
52Qwen 3.6 Plusmax61.9%$0.047–
53DeepSeek V4 Flash 0731high61.8%$0.012–
54MiMo V2.5 Proxhigh61.4%$0.037–
55DeepSeek V4max61.4%$0.149–
56Sonnet 4high61.2%$0.237–
57GPT-5.4 Nanohigh61.2%$0.016–
58MiniMax M2.7high61.2%$0.023–
59Grok 3 Minihigh61.1%$0.017–
60GPT-5high61.1%$0.086–
61Mistral Large 3none61.0%$0.024–
62GLM 4.5xhigh61.0%$0.038–
63GPT-5.4 Minixhigh60.9%$0.076–
64Gemini 2.5 Prohigh60.8%$0.095–
65Gemini 3.5 Flash-Litehigh60.7%$0.029–
66Haiku 4.5xhigh60.6%$0.089–
67Kimi K2medium60.6%$0.038–
68Sonnet 3.7medium60.4%$0.232–
69GPT-5 Minihigh60.2%$0.021–
70MiMo V2.5xhigh59.9%$0.012–
71Gemini 2.5 Flashmedium59.8%$0.026–
72o3high59.7%$0.133–
73Grok 3none59.7%$0.170–
74MiniMax M2.5high59.6%$0.024–
75Gemini 3.1 Flash-Litehigh59.4%$0.023–
76o4 minihigh59.0%$0.074–
77MiniMax M2.1–58.9%$0.022–
78Mistral Medium 3.5high58.8%$0.146–
79gpt-oss-120bhigh58.2%$0.0096–
80Laguna M.1–58.2%––
81GPT-4.1 Minihigh57.9%$0.026–
82Gemini 2.5 Flash Litenone57.6%$0.0086–
83GLM 4.6none56.8%$0.045–
84Laguna XS.2–56.3%––
85DeepSeek V3none54.7%$0.052–
86Trinity Large Thinking–54.7%$0.019–
87Nemotron 3.5 Lightning–54.2%$0.0048–
88DeepSeek R1xhigh54.1%$0.087–
89Claude 3.5 Sonnet Latest–53.6%$0.238–
90gpt-oss-20bhigh53.1%$0.0045–
91DeepSeek V3.1none51.5%$0.037–
92Grok 2none51.1%$0.115–
93DeepSeek V3.2high51.0%$0.038–
94Haiku 3.5medium50.8%$0.063–
95Mistral Medium 3.1none50.7%$0.024–
96Llama 4 Maverickmedium49.7%$0.014–
97Magistral Medium 1.2medium47.4%$0.125–
98Llama 4 Scoutmedium46.8%$0.012–
99GLM 4.7none46.4%$0.039–
100Command Anone46.0%$0.206–
101GPT-4ohigh45.9%––
102GPT-4o minihigh45.5%$0.0088–
103o3 minihigh45.3%$0.078–
104Mistral Small 3.1none44.2%$0.0044–
105Magistral Small 1.2–44.0%$0.030–
106Gemini 2.0 Pronone43.4%$0.105–
107GPT-4.1 Nanohigh42.1%$0.0065–
108Jamba 1.6 Largenone41.5%$0.150–
109Gemini 1.5 Pronone40.5%$0.131–
110Jamba 1.5 Largenone39.4%$0.149–
111Llama 3.1max38.9%$0.053–
112Gemini 1.5 Flashnone38.2%$0.0079–
113Jamba 1.6 Mininone38.0%$0.015–
114Jamba 1.5 Mininone33.9%$0.015–
115Gemini 2.0 Flashmedium33.7%$0.010–