Gemini 3.5 Flash Lite: benchmark results
Google's Gemini 3.5 Flash-Lite (July 21, 2026), a low-latency, cost-effective multimodal model for high-throughput subagent and document-parsing tasks, with a 1-million-token context. Provider: Google. Released 2026-07-21. Access: API.
Aggregate Rating 1724 ± 7, rank #156 of 1656 rated models, from 484 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Interpretive Canons (BVerfG) - Historical Interpretation | 72.7 | Positive-class F1 (%; historical interpretation (geschichte) | 100 |
| Polish Knowledge Benchmark - Media & Journalism | 88.68 | Accuracy (%) | 97.1 |
| Polish Knowledge Benchmark - Sport | 87.14 | Accuracy (%) | 97.1 |
| AI Chess Leaderboard (Continuation) | 1476 | Elo | 95.4 |
| Polish Knowledge Benchmark - Literature & Polish Language | 92.31 | Accuracy (%) | 94.1 |
| Polish Knowledge Benchmark - Music | 88.31 | Accuracy (%) | 94.1 |
| Polish Knowledge Benchmark - Theatre | 86.54 | Accuracy (%) | 94.1 |
| SvelteBench | 100 | Average pass@1 (%) | 93.7 |
| Polish Knowledge Benchmark - Film & Theatre | 91.35 | Accuracy (%) | 93.4 |
| Polish Knowledge Benchmark - Art | 88.5 | Accuracy (%) | 92.6 |
| Polish Knowledge Benchmark - Education & Science | 85.91 | Accuracy (%) | 92.6 |
| Polish Knowledge Benchmark - Geography | 88.68 | Accuracy (%) | 92.6 |
Interactive version: theaggregate.ai/model?slug=gemini-3-5-flash-lite · How It Works · Data refreshed daily, snapshot 2026-10-20.