Atlas

Benchmarks

← All benchmarks

Bench to the Future 3 — Binary

Forecasting · 2026-06-29

The binary component of BTF-3 contains 1,515 resolved forecasting questions researched against a frozen web corpus. Forecast probabilities are evaluated by mean Brier score, so lower scores are better.

Top models (lower is better)

ModelScore
Claude Opus 50.1
Opus 4.80.1
Claude Fable 50.1
GPT-5.6 Sol0.1
GPT-5.50.1
Sonnet 50.2
Loading Atlas data…