68.3%
#1
CorpFin v2 accuracy of Kimi K2.5, the highest of the top 15; Qwen 3 Max Thinking scored 68%
What the chart shows
CorpFin v2 asks models questions about credit agreements that can run past 200 pages, from extracting terms to numeric reasoning and cross-referencing sections. Kimi K2.5 led at 68.3%, just ahead of Qwen 3 Max Thinking (68%) and Claude Opus 4.6 (Thinking) (67%).
The top 15 models are separated by 5.2 percentage points, down to GPT 4.1 at about 63%, and none breaks 70%. The benchmark also varies how much of the document a model is given, from the exact pages to as much as fits in its context.
Key numbers
| Measure | Value |
|---|---|
| Kimi K2.5 | 68.26% |
| Qwen 3 Max Thinking | 68.03% |
| Claude Opus 4.6 (Thinking) | 67.02% |
| Grok 4 Fast (Reasoning) | 66.90% |
| GPT 4.1, 15th | 63.05% |
| Gap, 1st to 15th | 5.21 percentage points |
How to read it
- Questions were developed with financial analysts, lawyers and academics, per the AI Index.
- Scores are from Vals.ai’s CorpFin v2 leaderboard; the AI Index shows the top 15 models.
Source: Vals.ai, 2026, via the AI Index Report 2026 (Stanford HAI), figure 2.5.10: CorpFin v2: accuracy. Chart and text by AI Adoption.
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Discussion