Hedge-Bench - Risk Pass@1: leaderboard
Metric: Pass@1 (%): share of trials in the Risk task category that earned a perfect 4 of 4 rubric score; 102 open-ended financial-analyst research tasks derived from recorded expert analyst discussions, run in the Harbor harness with the Terminus 2 agent over a sandboxed document corpus, eight trials per task, graded by a Gemini-3.1-Pro judge against expert rubric moves with hallucination-supported moves discounted; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 12.4 |
| 2 | Gemini 3.5 Flash | 5.7 |
| 3 | Claude Opus 4.8 | 5.6 |
| 4 | Claude Opus 4.7 | 5.4 |
| 5 | GPT-5.5 | 1.8 |
| 6 | Gemini 3.1 Pro (Preview) | 1.7 |
| 7 | Claude Haiku 4.5 | 0.9 |
| 8 | GPT-5.4 Mini | 0 |
Interactive version: theaggregate.ai/benchmark?slug=hedge-bench-risk-pass-1 · How It Works · Data refreshed daily, snapshot 2026-09-29.