Gemini 3.1: benchmark results
Provider: Google. Access: API.
Aggregate Rating 1810 ± 39, rank #76 of 10916 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM-WikiRace | 64.93 | Average Success (%) | 100 |
| NVIDIA ComputeEval - CCCL | 91.3 | Pass@1 (%, zero-shot, 2026.1 release, 80 problems) | 97.4 |
| NVIDIA ComputeEval - cuda-runtime | 100 | Pass@1 (%, zero-shot, 2026.1 release, 29 problems) | 97.4 |
| NVIDIA ComputeEval | 69.1 | Pass@1 (%, zero-shot, 2026.1 release, 566 problems) | 94.7 |
| NVIDIA ComputeEval - cuda-kernels | 86.8 | Pass@1 (%, zero-shot, 2026.1 release, 114 problems) | 94.7 |
| HumanCLAW-Bench - Interact | 16.8 | Success Rate (%) | 90 |
| HumanCLAW-Bench - Navigate | 42.4 | Success Rate (%) | 90 |
| NVIDIA ComputeEval - cuDNN | 25.7 | Pass@1 (%, zero-shot, 2026.1 release, 113 problems) | 89.5 |
| NVIDIA ComputeEval - cuBLAS | 96.3 | Pass@1 (%, zero-shot, 2026.1 release, 81 problems) | 86.8 |
| HumanCLAW-Bench - Find | 64.9 | Success Rate (%) | 80 |
| NVIDIA ComputeEval - Math Libs | 82.2 | Pass@1 (%, zero-shot, 2026.1 release, 101 problems) | 73.7 |
| AttuneBench | 52.9 | Composite (self-reported) | 60 |
Interactive version: theaggregate.ai/model?slug=gemini-3-1 · How It Works · Data refreshed daily, snapshot 2026-11-08.