63.3%
#1
Finance Agent v1.1 accuracy of Claude Sonnet 4.6, the highest of the top 15; Kimi K2.5 was 15th at 50.6%
What the chart shows
Finance Agent v1.1 tests AI agents on 537 questions typical of an entry-level financial analyst’s work, such as retrieving information, market research and financial projections. Claude Sonnet 4.6 led at 63.3%, ahead of Claude Opus 4.6 (Thinking) at 60% and Gemini 3.1 Pro Preview at 59.7%.
Results are more spread out than on the other finance benchmarks: the top 15 span 12.7 percentage points, down to Kimi K2.5 at 50.6%. Even the leader falls short of two-thirds accuracy.
Key numbers
| Measure | Value |
|---|---|
| Claude Sonnet 4.6 | 63.33% |
| Claude Opus 4.6 (Thinking) | 60.05% |
| Gemini 3.1 Pro Preview (02/26) | 59.72% |
| GPT 5.2 | 58.53% |
| Kimi K2.5, 15th | 50.62% |
| Gap, 1st to 15th | 12.71 percentage points |
How to read it
- The benchmark was developed with Stanford researchers, a global systemically important bank and industry experts, per the AI Index.
- Scores are from Vals.ai’s Finance Agent v1.1 leaderboard; the AI Index shows the top 15 models.
Source: Vals.ai, 2026, via the AI Index Report 2026 (Stanford HAI), figure 2.5.11: Finance Agent v1.1: 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