Bench to the Future 3 — September Panel — Pooled
Forecasting
BTF-3 September 2026 panel: 2,386 resolved questions from June-August 2026 researched against a frozen web corpus. The pooled score combines 2,015 binary Brier scores and 371 numeric ranked probability scores, weighting each numeric question three times as much as each binary question. This panel is distinct from the earlier 1,907-question release.
Top models (lower is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 0.1 |
| Opus 4.8 | 0.1 |
| GPT-6 Astra | 0.1 |
| GPT-5.6 Sol | 0.1 |
| Muse Spark 1.3 | 0.1 |
| Sonnet 5 | 0.1 |
| GLM-5.3 Flash | 0.1 |
| GLM-5.3 | 0.1 |