Google DeepMind
gemini-3-pro-preview
Made by Google DeepMind. Released 2025-11-18. Accessibility not recorded. We hold no price for this model: no seller we can read lists it any more, so what follows is what was measured, not what it cost.
What it scores, by house
| Benchmark | House | Score | Evaluated | Scaffold |
|---|---|---|---|---|
| chess puzzles | Epoch AI | 0.3±0.05 | 2025-12-08 | — |
| frontiermath | Epoch AI | 0.4±0.03 | 2025-11-21 | — |
| frontiermath tier 4 | Epoch AI | 0.2±0.06 | 2025-11-21 | — |
| gpqa diamond | Epoch AI | 0.9±0.02 | 2025-11-19 | — |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.9±0.04 | 2025-11-19 | — |
| swe bench verified | Epoch AI | 0.7±0.02 | 2026-02-13 | — |
| ale bench | collected by Epoch AI | 1177 | — | — |
| algotune | collected by Epoch AI | 1.8 | — | — |
| apex agents | collected by Epoch AI | 0.5 | — | — |
| arc aginear its ceiling | collected by Epoch AI | 0.8 | — | — |
| arc agi 2 | collected by Epoch AI | 0.3 | — | — |
| balrog | collected by Epoch AI | 0.6 | — | — |
| cl bench | collected by Epoch AI | 0.2 | — | — |
| critpt | collected by Epoch AI | 0.1 | — | — |
| deepresearchbench | collected by Epoch AI | 0.5 | — | low |
| enigma eval | collected by Epoch AI | 0.2 | — | — |
| forecastbench | collected by Epoch AI | 61.2 | — | — |
| gdpval | collected by Epoch AI | 0.4 | — | — |
| geobench | collected by Epoch AI | 4145 | — | — |
| gso | collected by Epoch AI | 0.2 | — | OpenHands |
| hle | collected by Epoch AI | 0.4 | — | — |
| metr time horizons | collected by Epoch AI | 224 | — | — |
| proofbench | collected by Epoch AI | 0.2 | — | — |
| rli | collected by Epoch AI | 0.0 | — | — |
| simplebench | collected by Epoch AI | 0.8 | — | — |
| terminalbench | collected by Epoch AI | 0.6 | — | Droid |
| terminalbench | collected by Epoch AI | 0.6 | — | CodeBrain-1 |
| terminalbench | collected by Epoch AI | 0.7 | — | Ante |
| terminalbench | collected by Epoch AI | 0.6 | — | Letta Code |
| terminalbench | collected by Epoch AI | 0.6 | — | Terminus 2 |
| terminalbench | collected by Epoch AI | 0.6 | — | CodeBrain-1.5 |
| terminalbench | collected by Epoch AI | 0.7 | — | SageAgent |
| terminalbench | collected by Epoch AI | 0.6 | — | II-Agent |
| vending bench 2 | collected by Epoch AI | 5478 | — | — |
| vpct | collected by Epoch AI | 0.9 | — | — |
| webdev arena | collected by Epoch AI | 1438 | — | — |
| weirdml | collected by Epoch AI | 0.7 | — | — |
These scores are not comparable between houses and are not added up. A score is dated by when it was evaluated, not when the model was released, and a score without its scaffold is not a measurement — which is why both are printed.