Google DeepMindGemini
Gemini 3.5 Flash
Google DeepMind · Gemini. Released May 19, 2026.
Gemini 3.5 Flash is a model from Google DeepMind in the Gemini family, released May 19, 2026. evals.report tracks 29 reported Gemini 3.5 Flash benchmark scores across FrontierMath, DeepSWE, ARC-AGI-1, ARC-AGI-2, LMArena, LiveBench, GPQA Diamond, Humanity's Last Exam, and 21 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.
29 results
Benchmark results 29
Compare this model| Benchmark | Category | Score | Metric | Status | Date | |
|---|---|---|---|---|---|---|
| FrontierMath | Reasoning | 38.97% | accuracy | Official | May 19, 2026 | Details |
| DeepSWE | Coding | 28.32% | % resolved | Official | May 19, 2026 | Details |
| ARC-AGI-1 | Reasoning | 92.5% | accuracy | Official | May 19, 2026 | Details |
| ARC-AGI-2 | Reasoning | 72.08% | accuracy | Official | May 19, 2026 | Details |
| LMArena | Chat preference | 1482 | source-defined rating | Official | May 19, 2026 | Details |
| LiveBench | Reasoning | 75.02% | score | Official | May 19, 2026 | Details |
| GPQA Diamond | Reasoning | 92.8% | accuracy | Official | May 19, 2026 | Details |
| Humanity's Last Exam | Reasoning | 42.5% | accuracy | Official | May 19, 2026 | Details |
| AIME (OTIS Mock) | Reasoning | 95.6% | accuracy | Official | May 19, 2026 | Details |
| SimpleQA Verified | Other | 68.4% | accuracy | Official | May 19, 2026 | Details |
| GBA Eval | Coding | 6.7% | overall score | Official | May 19, 2026 | Details |
| WeirdML | Coding | 62.6% | average accuracy | Official | May 19, 2026 | Details |
| CursorBench | Agents | 49.8% | score | Official | May 19, 2026 | Details |
| Artificial Analysis Intelligence Index | Reasoning | 55.3 | Index | Unverified | May 19, 2026 | Details |
| Epoch Capabilities Index | Reasoning | 156.3 | Index | Official | May 19, 2026 | Details |
| OSWorld | Agents | 78.4% | task success rate | Unverified | May 19, 2026 | Details |
| τ²-bench (Telecom) | Tool use | 95.3% | pass^1 | Official | May 19, 2026 | Details |
| AIME 2026 | Reasoning | 95.00% | accuracy | Official | May 19, 2026 | Details |
| GDPval | Agents | 1659 | Elo | Official | May 19, 2026 | Details |
| SciCode | Coding | 53.1% | accuracy | Unverified | May 19, 2026 | Details |
| AA-Omniscience: Knowledge and Hallucination Benchmark | Reasoning | 23 | AA-Omniscience Index | Official | May 19, 2026 | Details |
| IFBench | Reasoning | 76.3% | accuracy | Official | May 19, 2026 | Details |
| OpenAI-MRCR v2 (Multi-Round Coreference Resolution) | Reasoning | 26.6% | accuracy (mean SequenceMatcher similarity) | Verified | May 19, 2026 | Details |
| WebDev Arena | Chat preference | 1506 | Elo | Verified | May 19, 2026 | Details |
| Design Arena | Chat preference | 1297 | Elo | Verified | May 19, 2026 | Details |
| CharXiv | Multimodal | 84.2% | accuracy | Unverified | May 19, 2026 | Details |
| MathArena HMMT February 2026 | Reasoning | 95.45% | accuracy | Official | May 19, 2026 | Details |
| FrontierMath Tier 4 | Reasoning | 14.6% | accuracy | Official | May 19, 2026 | Details |
| Vibe Code Bench | Coding | 48.68% | Overall accuracy | Verified | May 19, 2026 | Details |