Gemini 2.5 Flash (Thinking): benchmark results
Gemini 2.5 Flash evaluated with thinking enabled. Provider: Google. Released 2025-06-17. Access: API.
Aggregate Rating 1632 ± 9, rank #590 of 2168 rated models, from 401 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Cedar - Multimodal - Arabic | 45.75 | Accuracy (%) | 100 |
| Cedar - Multimodal - Hindi | 47.73 | Accuracy (%) | 100 |
| Cedar - Multimodal - Swahili | 42.6 | Accuracy (%) | 100 |
| Cedar - Text-Only - Arabic | 51.48 | Accuracy (%) | 100 |
| Cedar - Text-Only - Hindi | 51.27 | Accuracy (%) | 100 |
| Cedar - Text-Only - Spanish | 52.98 | Accuracy (%) | 100 |
| EDCT-Bench (Binary) | 60 | Counterfactual Consistency Score, binary (%; mean over cited | 100 |
| EDCT-Bench (Rubric-Graded) | 60 | Counterfactual Consistency Score, rubric-graded (%; LLM-judg | 100 |
| EuroEval Catalan | 68.12 | Average Score (%) | 100 |
| EuroEval Slovene NLU | 63.65 | NLU Average Score (%) | 100 |
| IHBench - Recovery Quality | 70.4 | Recovery pass rate (%; share of 428 interruption points meet | 100 |
| IHBench - Recovery Quality - Filler | 68 | Recovery pass rate (%; 60 backchannel fillers, where the tur | 100 |
Interactive version: theaggregate.ai/model?slug=gemini-2-5-flash-thinking · How It Works · Data refreshed daily, snapshot 2026-10-20.