FutureEval: leaderboard
Metaculus benchmark for AI forecasting accuracy on unresolved real-world questions across science, technology, health, geopolitics, AI, and other domains. Models make probabilistic forecasts and are ranked by a unified log-score-based forecasting score against community and pro forecaster baselines.
Metric: Unified Forecasting Score. Source: www.metaculus.com. Status: saturation imminent. 93 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 (High) | 15.41 |
| 2 | Claude Fable 5 (High) | 13.77 |
| 3 | Claude Opus 4.8 (High) | 12.97 |
| 4 | Gemini 3.6 Flash | 12.7 |
| 5 | GPT-5.5 Instant | 12.66 |
| 6 | GPT-5.6 Sol (High) | 12.45 |
| 7 | Gemini 3.1 Pro (Preview) (High) | 12.23 |
| 8 | GPT-5.1 (High) | 12 |
| 9 | GPT-5.5 (High) | 11.95 |
| 10 | O3 | 11.73 |
| 11 | GPT-5 | 11.49 |
| 12 | GPT-5 (High) | 11.47 |
| 13 | Kimi K3 | 11.31 |
| 14 | Gemini 3.5 Flash | 11.29 |
| 15 | Claude Sonnet 5 | 11.18 |
Interactive version: theaggregate.ai/benchmark?slug=futureeval · How It Works · Data refreshed daily, snapshot 2026-09-05.