FinRiskAtlas - Evidence-Grounded Processing - Institution Matching: leaderboard
Metric: Accuracy (%). Source: arxiv.org. 33 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 93.6 |
| 2 | Qwen3-Next-80B | 90.8 |
| 3 | GPT-5.6 Terra | 90.4 |
| 4 | Qwen 3 32B | 90 |
| 5 | GPT-5.6 Sol | 90 |
| 6 | Gemini 2.5 Flash | 89.2 |
| 7 | DeepSeek V4 Flash | 88 |
| 8 | Qwen 3 235B A22B | 88 |
| 9 | GLM-5 | 87.6 |
| 10 | Kimi K2.5 | 87.2 |
| 11 | DeepSeek V4 Pro | 86.8 |
| 12 | Seed 2.0 Mini | 86.8 |
| 13 | GLM-5.2 | 85.6 |
| 14 | GPT-5.6 Luna | 85.6 |
| 15 | Kimi K3 | 84.8 |
Interactive version: theaggregate.ai/benchmark?slug=finriskatlas-evidence-grounded-processing-institution-matching · How It Works · Data refreshed daily, snapshot 2026-09-19.