ForecastBench Baseline - Scratchpad - Overall: leaderboard
ForecastBench Baseline leaderboard, where the ForecastBench team runs models out of the box, scratchpad prompt (step-by-step reasoning before the probability). Overall index, computed from the averaged dataset and market Brier scores rather than the mean of the two indices; difficulty-adjusted Brier Index on 0-100 (higher is better; 50 equals always forecasting 0.5). Superforecaster and public medians, naive baselines and LLM crowds are excluded.
Metric: Overall Brier Index (difficulty-adjusted, higher is better). Source: www.forecastbench.org. Saturation forecast: Estimated already saturated. 36 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 (2025-04-16) | 61.4 | #125 |
| 2 | O4 Mini (2025-04-16) | 60.6 | #183 |
| 3 | Claude 3.7 Sonnet (20250219) | 60.6 | #209 |
| 4 | Claude Opus 4.1 (20250805) | 60.4 | #139 |
| 5 | GPT-4.1 | 60.2 | #258 |
| 6 | Claude Opus 4 (20250514) | 60 | #164 |
| 7 | GPT-4.5 (Preview) | 60 | #260 |
| 8 | Claude 3.5 Sonnet (20241022) | 59.6 | #302 |
| 9 | Gemini 2.5 Flash (Preview 04-17) | 59.3 | #246 |
| 10 | DeepSeek R1 | 58.8 | #262 |
| 11 | Claude 3.5 Sonnet (20240620) | 58.3 | #308 |
| 12 | Gemini 1.5 Pro | 58.1 | #454 |
| 13 | Claude Sonnet 4 (20250514) | 58.1 | #223 |
| 14 | DeepSeek V3 | 57.9 | #329 |
| 15 | QwQ 32B-Preview | 57.3 | #671 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=forecastbench-baseline-scratchpad-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.