ParanoiaEval - Task Success: leaderboard
Metric: Task success rate (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Sol (High) | 97.6 |
| 2 | Claude Opus 5 (High) | 95.8 |
| 3 | GPT-5.6 Terra (High) | 93.5 |
| 4 | Claude Sonnet 5 (High) | 93.4 |
| 5 | GPT-5.6 Luna (High) | 92.8 |
| 6 | GPT-5.5 (High) | 92.2 |
| 7 | Claude Sonnet 4.6 (High) | 90.3 |
| 8 | Claude Haiku 4.5 (High) | 88.3 |
Interactive version: theaggregate.ai/benchmark?slug=paranoiaeval-task-success · How It Works · Data refreshed daily, snapshot 2026-10-09.