Gemini 3.1 Flash Lite (Minimal): benchmark results
Provider: Google. Released 2026-05-07. Access: API.
Aggregate Rating 1627 ± 27, rank #616 of 2170 rated models, from 28 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SteerBench-Work | 92.8 | Mean Trial Accuracy (%) | 100 |
| SteerBench-Work - Pass^5 | 91.5 | Scenarios Correct in All 5 Trials (%) | 100 |
| PRISM-VLM - Multi-Question Robustness | 53 | Per-axis mean score (%; share of items whose 3-5 bundled que | 97.6 |
| PRISM-VLM - Quality | 71.3 | Per-axis mean score (%; base accuracy on the recycled questi | 97.6 |
| PRISM-VLM | 67.5 | PScore (%; harmonic mean of the seven per-axis mean scores, | 97.5 |
| PRISM-VLM - Visual Discrimination | 92.6 | Per-axis mean score (%; forced A/B choice between the gold c | 97.5 |
| PRISM-VLM - Counting | 62.8 | Per-axis mean score (%; exact-integer accuracy on synthesize | 92.7 |
| PRISM-VLM - Hallucination Resistance | 73.1 | Per-axis mean score (%; share of trap questions (absent enti | 90.2 |
| PRISM-VLM - Instruction Following | 83.6 | Per-axis mean score (%; mean compliance with 1-3 verifiable | 87.8 |
| Wolfram LLM Benchmarking Project | 52.8 | Correct Functionality (%) | 68.2 |
| DGEval - Regulatory Recall | 17.8 | Accuracy (%) | 67.6 |
| K-Bench | 98.13 | Overall (%, Default v1 prompt) | 67.2 |
Interactive version: theaggregate.ai/model?slug=gemini-3-1-flash-lite-minimal · How It Works · Data refreshed daily, snapshot 2026-11-05.