gemini-robotics-er-1.5-preview ranks first on Point-Bench.
Google DeepMind Released September 26th, 2025 Family Gemini Robotics The lab's model card Compare with others Google DeepMind Sep 2025 Closed weights Mid-size
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 83.3% | #97 of 302 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Aider Polyglot | 57.3% | #16 of 30 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Point-Bench | 67.1% | #1 of 17 | Official leaderboard | |
| MMMU | 80.7% | #18 of 103 | Lab model card | |
| MVPBench | 72.5% | #1 of 4 | Lab model card | |
| VPCT | 40.8% | #12 of 27 | VPCT leaderboard, via Epoch AI | |
| RoboRewardBench | 0.91 | #10 of 22 | Official leaderboard | |
| Where2Place | 59% | #2 of 4 | Lab model card | |
| CVBench | 84.3% | #8 of 15 | Lab model card | |
| ClockBench | 15% | #22 of 37 | Official leaderboard | |
| ERQA | 54.8% | #23 of 36 | Lab model card | |
| BLINK | 57.8% | #22 of 29 | Lab model card | |
| RefSpatial-Bench | 48.5% | #13 of 16 | Lab model card | |
| EmbSpatialBench | 78.4% | #14 of 16 | Lab model card | |
| VSI-Bench | 45.8% | #7 of 7 | Lab model card | |
| MindCube | 54.7% | #4 of 4 | Lab model card | |
| OpenEQA | 55.0 | #2 of 2 | Lab model card | |
| SAT (Spatial Aptitude Training) | 76.7% | #1 of 1 | Lab model card |
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| Point-Bench | 71.6% Point-Bench (Table 19, thinking on). Thinking off 73.3 | 67.1% Leaderboard | +4.5 pts |
Every result in the longer documents the lab published with it. 16 of them aren't in the launch post.