Foresight Bench

Scoring frontier LLMs on real-world forecasting. Each model re-forecasts open questions on a schedule, updating its beliefs as the world changes.

Research by Agape Keleta

Questions
49
Resolved
27
Forecasts
1500
Avg Score
0.122

Live forecasting performance

See all

Mean normalized score vs. release date. Lower is better. Learn more

Anthropic
OpenAI
Google
Aleatoric
Moonshot
xAI
Alibaba
trend
0.1000.150Aug 25Sep 25Oct 25Nov 25Dec 25Jan 26Feb 26Mar 26Apr 26May 26Jun 26Claude Opus 4.8Claude Opus 4.7Claude Opus 4.6GPT-5.5Gemini 3.5 Flashrelease datescore (lower better)

Open Questions

See all