OptionWorld - Risk Hypothesis: leaderboard
OptionWorld (2026): three runs over192targets,24trajectories per system,60-day evolving environments. API routes use fresh daily conversations,4096completion tokens and16turns/day; temperature/seed omitted. The separate local Astra bridge has only gpt-6-astra-pro-local identity and is not merged into a public release. Rate (%).
Metric: Rate (%). Source: arxiv.org. Saturation forecast: Around 2030. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro (0813) | 86.3 | #52 |
| 2 | Qwen 3.8 Max (0902) | 63 | #28 |
| 3 | Claude Fable 5.1 | 60.2 | #9 |
| 4 | Grok 4.6 | 58.9 | #26 |
| 5 | Gemini 3.1 Pro (Preview) | 56.2 | #60 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=optionworld-risk-hypothesis · How It Works · Data refreshed daily, snapshot 2026-10-22.