← All models

Gemini Deep Think 3 (December 2025 evaluation)

Google

This benchmark entry preserves MathArena's Gemini Deep Think 3 (12/25) Putnam evaluation, recorded with the restricted endpoint models/gdm-eval-model-bcn. The source identifies an earlier evaluation snapshot but does not establish that it is identical to the December 4 consumer rollout. It is distinct from the February 12, 2026 Deep Think upgrade; the evaluator's November 15 date is not adopted as an official public release date.

Benchmark scores