Proof of Time - 15-Message Budget: leaderboard
The Proof of Time forecasting tasks and offline tool-using agent with the interaction limit cut to 15 messages. Sample-weighted exact-match accuracy (%), higher is better (Proof of Time, Ye, Chen et al., Harvard University and Mass General Brigham, 2026).
Metric: Sample-Weighted Accuracy (%), 15-message agent budget. Source: arxiv.org. Saturation forecast: Not forecast. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 27.2 | #111 |
| 2 | Gemini 2.5 Pro | 26.7 | #262 |
| 3 | GPT-5.1 | 26.7 | #225 |
| 4 | GPT-5 Mini | 22.3 | #349 |
| 5 | GPT-5 Nano | 21.7 | #888 |
| 6 | Gemini 3 Flash (Preview) | 20.3 | #141 |
| 7 | GPT-5.2 | 17.9 | #172 |
| 8 | Gemini 2.5 Flash | 17.2 | #452 |
| 9 | Claude Opus 4.5 | 16.7 | #152 |
| 10 | Claude Haiku 4.5 | 8.4 | #568 |
| 11 | Claude Sonnet 4.5 | 2.9 | #244 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=proof-of-time-15-message-budget · How It Works · Data refreshed daily, snapshot 2026-11-12.