Gemini 2.5 Flash Lite: benchmark results
Google's most cost-efficient Gemini 2.5 tier for low-latency workloads. Provider: Google. Released 2025-06-17. Access: API.
Aggregate Rating 1606 ± 5, rank #422 of 1656 rated models, from 1582 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| InfiniteBM Heads-Up No-Limit Hold'em | 1684.29 | Arena Elo (elo) | 100 |
| KhatianDoc - Document QA EM | 21.67 | Exact match (%; refusal-normalized, 300-item stratified docu | 100 |
| KoALa-Bench - CLIcK | 67.27 | Accuracy (%) on the spoken short-form CLIcK multiple-choice | 100 |
| KoALa-Bench - KCSAT | 81.18 | Accuracy (%) on the long-form listening questions of the Kor | 100 |
| PlantMicro - Organelle | 56.48 | Organelle identification accuracy (%); four-option multiple | 100 |
| RAI-Bench - RAG Robustness (HY Abstention) | 93 | Rate (%) | 100 |
| RHELM (Long Context) - Hallucination | 7.6 | Accuracy (%; 197 hallucination questions, LLM judge, dialogu | 100 |
| RHELM (Long Context) - Mixed | 13.1 | Accuracy (%; 210 mixed dialogue-and-document questions, LLM | 100 |
| EuroEval Danish NLU - ScaLA DA | 74.06 | Matthews correlation coefficient (%) | 98.7 |
| Vectara Hallucination Leaderboard | 96.7 | Factual Consistency Rate (%) | 98.1 |
| EuroEval Lithuanian NLU - Atsiliepimai | 50.18 | Matthews correlation coefficient (%) | 98 |
| Open Portuguese LLM - OAB Exams | 67.84 | Accuracy (%) | 97.8 |
Interactive version: theaggregate.ai/model?slug=gemini-2-5-flash-lite · How It Works · Data refreshed daily, snapshot 2026-10-20.