FinRiskAtlas - Applied Review - Legal-Outcome Prediction: leaderboard
FinRiskAtlas fixed-evidence review operation: predicting a case disposition from facts and procedural record with the outcome withheld, scored by accuracy (FinRiskAtlas, arXiv 2608.25325v1, Appendix Table 9). Percent; higher is better. Reported for 33 archived model configurations under zero-shot direct-answer inference.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around 2030. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K3 | 68.15 | #35 |
| 2 | GPT-5.6 Terra | 67.83 | #40 |
| 3 | GPT-5.4 | 67.51 | #83 |
| 4 | GPT-5.6 Sol | 67.09 | #22 |
| 5 | Qwen 3.7 Max | 66.03 | #85 |
| 6 | Kimi K2.5 | 65.93 | #152 |
| 7 | GPT-5.6 Luna | 65.93 | #79 |
| 8 | GLM-5.2 | 65.71 | #80 |
| 9 | Qwen3-Next-80B | 65.71 | #321 |
| 10 | Claude Opus 4.7 | 65.19 | #51 |
| 11 | Qwen 3 235B A22B | 65.08 | #323 |
| 12 | DeepSeek V4 Pro | 64.97 | #112 |
| 13 | Qwen 3 14B | 64.97 | #542 |
| 14 | GPT-5.4 Mini | 64.97 | #228 |
| 15 | GLM-5 | 64.76 | #149 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=finriskatlas-applied-review-legal-outcome-prediction · How It Works · Data refreshed daily, snapshot 2026-11-05.