Gemini 3.1 Flash Lite (Medium): benchmark results
Provider: Google. Released 2026-05-07. Access: API.
Aggregate Rating 1628 ± 15, rank #612 of 2170 rated models, from 38 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMGist - Diagram and OCR | 73.4 | Accuracy (%) on the Diagram and OCR dimension (items kept fr | 96.2 |
| SLR-Bench - Basic | 100 | Accuracy (%) | 86.4 |
| DGEval - IMDG Code MCQ | 79.2 | Accuracy (%) | 82.4 |
| MMGist - Medical | 41.8 | Accuracy (%) on the Medical dimension (items kept from SLAKE | 80.8 |
| MMGist - Visual Perception | 62.4 | Accuracy (%) on the Visual Perception dimension (items kept | 80.8 |
| MMGist - Expert Knowledge | 54.1 | Accuracy (%) on the Expert Knowledge dimension (items kept f | 76.9 |
| DGEval - Regulatory Recall | 19.3 | Accuracy (%) | 73.5 |
| MMGist - Visual Logic | 27.2 | Accuracy (%) on the Visual Logic dimension (items kept from | 69.2 |
| NVIDIA ComputeEval - CCCL | 82.5 | Pass@1 (%, zero-shot, 2026.1 release, 80 problems) | 65.8 |
| MMGist - STEM Reasoning | 41.5 | Accuracy (%) on the STEM Reasoning dimension (items kept fro | 65.4 |
| SLR-Bench - Hard | 6 | Accuracy (%) | 63.6 |
| ObviousBench | 95.83 | Answer pass³ (%) | 61.9 |
Interactive version: theaggregate.ai/model?slug=gemini-3-1-flash-lite-medium · How It Works · Data refreshed daily, snapshot 2026-11-05.