ForecastBench: leaderboard
Measures LLM ability to forecast future events - think prediction markets. 1,000 auto-generated questions from Polymarket, Manifold, and real-world data. Compares against superforecasters and the public.
Metric: Overall Score (higher is better). Source: www.forecastbench.org. Source: www.forecastbench.org. Saturation forecast: Around 2031. 297 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.7 Flash (High) | 66.1 | #23 (Gemini 3.7 Flash) |
| 2 | Claude Fable 5 (Max) | 66 | #16 (Claude Fable 5) |
| 3 | Gemini 3.6 Flash (Medium) | 65.9 | #51 (Gemini 3.6 Flash) |
| 4 | Claude Fable 5 (High) | 65.8 | #16 (Claude Fable 5) |
| 5 | Kimi K3 (Max) | 65.6 | #30 (Kimi K3) |
| 6 | GPT-5 | 65.5 | #98 |
| 7 | Gemini 3 Pro (Preview) | 65.4 | #66 |
| 8 | Claude Fable 5 | 65.3 | #16 |
| 9 | Grok 4 | 65.1 | #181 |
| 10 | Grok 4 Fast (Reasoning) | 65.1 | #254 (Grok 4 Fast) |
| 11 | Claude Opus 4.5 (20251101) | 65 | #74 |
| 12 | GPT-5.6 Sol | 64.9 | #19 |
| 13 | Grok 4.1 Fast (Reasoning) | 64.9 | #217 (Grok 4.1 Fast) |
| 14 | GPT-4.5 (Preview) | 64.9 | #244 |
| 15 | Gemini 3 Flash (Preview) | 64.8 | #82 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=forecastbench · How It Works · Data refreshed daily, snapshot 2026-10-20.