Gemini 3 Flash (Low): benchmark results
Provider: Google. Released 2025-12-17. Access: API.
Aggregate Rating 1684 ± 18, rank #406 of 2169 rated models, from 42 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BARISTA - Phrase Grounding | 53.4 | COCO-style mAP (%, IoU 0.5 to 0.95, times 100) of the boxes | 100 |
| BARISTA - Referring Expression Generation | 0.76 | Correctness of the generated description of a marked region, | 100 |
| LegalScope - Overall Human | 83 | Score (%) | 100 |
| VocalGrad - Clarity | 54 | Accuracy (%) | 100 |
| VocalGrad - Pitch | 51.5 | Accuracy (%) | 100 |
| LegalScope - Case Human | 85.4 | Score (%) | 96.3 |
| VSTAT - Sequence | 61.6 | Score (%) on questions about sequence state structures: 1,50 | 95.5 |
| LegalScope - Exam Human | 80.6 | Score (%) | 92.6 |
| Japanese G2P (Parse Mode) | 0.94 | Kana character error rate (%; can exceed 100) in parse mode | 91.7 |
| VSTAT | 39.8 | Average score (%) over all questions: 1,500 questions on 834 | 90.9 |
| VSTAT - Attribute | 52.2 | Score (%) on questions about attribute-type state elements: | 90.9 |
| VSTAT - Count | 33.4 | Score (%) on questions about count-type state elements: 1,50 | 90.9 |
Interactive version: theaggregate.ai/model?slug=gemini-3-flash-low · How It Works · Data refreshed daily, snapshot 2026-10-23.