Will the highest Chess Puzzles score score in Epoch AI's benchmark data reach at least 82.0 % for a model released on or before 2026-12-13?
échéance 2026-12-13
confiance moyenne
preuve niveau 1
21 %estimation PolySignal
Critère de résolution
Resolves YES if, in a snapshot of Epoch AI's benchmark data (https://epoch.ai/data/benchmark_data.zip, file "chess_puzzles.csv", column "mean_score") taken on or after 2026-12-27, at least one model with a release date on or before 2026-12-13 has a value >= 82.0 %. Reference at creation: 72.0 % (gpt-6-astra_max, 2026-09-03). Voided if Epoch AI revises the methodology so that values are no longer comparable, or retires the file. Data: Epoch AI, CC BY 4.0.
État de départ vérifié
Computed base rate: 16% of past 60-day windows in Epoch AI's benchmark data saw the running record rise by at least the required amount (threshold 0.82, current record 0.72). Data: Epoch AI, CC BY 4.0, epoch.ai/benchmarks. Overlapping windows: small effective sample.
Les arguments ci-dessous ont été produits par le modèle en anglais. Ils sont affichés tels quels plutôt que traduits automatiquement, pour que rien sur cette page ne soit un texte que PolySignal n'a pas réellement écrit.
Ce qui fait monter
- The current record was set recently by a frontier-class model (gpt-6-astra_max), indicating an active release cycle [1].
- Recent advancements in fine-tuning for reasoning tasks like IMO/IOI suggest that specialized models can achieve significant performance leaps [1].
- Reported rapid declines in the cost of AI performance suggest a high rate of underlying architectural or hardware efficiency gains [0].
Ce qui fait descendre
- A 10-point improvement in a 60-day window is historically rare, occurring only 16% of the time [1].
- There are no confirmed reports or tier 1/2 announcements of models currently exceeding the 72% threshold [1].
Ce qui ferait le plus bouger ce chiffre
The timing of the next major frontier model release from leading AI labs before the December 13 deadline.
Comment ce chiffre est fabriqué
| Échéance | 2026-12-13 |
| Taux de base | 16% |
| Meilleure preuve | niveau 1 · 2 article(s) utilisé(s) |
| Estimations | 20 · 22 (écart 2 pts) |
| Confiance | moyenne |
| Plafond de preuve | non déclenché |
| Révisée le | 2026-10-14 09:00 |
| Modèle | gemini-3-flash-preview |
Sources consultées