Igor Kotenkov
It is great at writing - i'm using it to this day. It was good in one-shotting front-end. But agentic? … in my memory it was never a catch up in the most important and money making areas
anecdotalhigh-signal user
Google DeepMindGemini
Gemini 3 Pro
Google DeepMind · Gemini. Released Nov 18, 2025.
Gemini 3 Pro is a model from Google DeepMind in the Gemini family, released Nov 18, 2025. evals.report tracks 52 reported Gemini 3 Pro benchmark scores across FrontierMath, Berkeley Function Calling Leaderboard, ARC-AGI-1, LMArena, LiveBench, LiveCodeBench Pro, SWE-bench Pro, GPQA Diamond, and 44 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.
52 results
Benchmark results 52
Compare this model| Benchmark | Category | Score | Metric | Status | Date | |
|---|---|---|---|---|---|---|
| FrontierMath | Reasoning | 37.6% | accuracy | Official | Nov 18, 2025 | Details |
| Berkeley Function Calling Leaderboard | Tool use | 72.51% | accuracy | Official | Nov 18, 2025 | Details |
| ARC-AGI-1 | Reasoning | 75% | accuracy | Official | Nov 18, 2025 | Details |
| LMArena | Chat preference | 1479 | source-defined rating | Official | Nov 18, 2025 | Details |
| LiveBench | Reasoning | 73.39% | score | Official | Nov 18, 2025 | Details |
| LiveCodeBench Pro | Coding | 2439 | Codeforces Elo | Official | Nov 18, 2025 | Details |
| SWE-bench Pro | Coding | 43.30% | % resolved | Official | Nov 18, 2025 | Details |
| GPQA Diamond | Reasoning | 92.6% | accuracy | Official | Nov 18, 2025 | Details |
| SWE-bench Verified | Coding | 72.9% | % resolved | Official | Nov 18, 2025 | Details |
| Humanity's Last Exam | Reasoning | 38.3% | accuracy | Official | Nov 18, 2025 | Details |
| MMMU-Pro | Multimodal | 81.0% | accuracy | Official | Nov 18, 2025 | Details |
| AIME (OTIS Mock) | Reasoning | 91.4% | accuracy | Official | Nov 18, 2025 | Details |
| SimpleQA Verified | Other | 72.9% | accuracy | Official | Nov 18, 2025 | Details |
| ARC-AGI-2 | Reasoning | 31.11% | accuracy | Official | Nov 18, 2025 | Details |
| WeirdML | Coding | 69.9% | average accuracy | Official | Nov 18, 2025 | Details |
| MCP Atlas | Tool use | 54.1% | pass rate | Official | Nov 18, 2025 | Details |
| PostTrainBench | Agents | 18.12% | weighted average score | Official | Nov 18, 2025 | Details |
| Artificial Analysis Intelligence Index | Reasoning | 48.4 | Index | Unverified | Nov 18, 2025 | Details |
| Epoch Capabilities Index | Reasoning | 153.4 | Index | Official | Nov 18, 2025 | Details |
| SWE-rebench | Coding | 56.5% | Resolved rate (pass@1) | Official | Nov 18, 2025 | Details |
| MMLU-Pro | Reasoning | 89.8% | accuracy | Verified | Nov 18, 2025 | Details |
| GAIA: A Benchmark for General AI Assistants | Agents | 38.5% | accuracy | Unverified | Nov 18, 2025 | Details |
| τ²-bench (Telecom) | Tool use | 85.4% | pass^1 | Unverified | Nov 18, 2025 | Details |
| AIME 2026 | Reasoning | 91.67% | accuracy | Official | Nov 18, 2025 | Details |
| GDPval | Agents | 1184 | Elo | Official | Nov 18, 2025 | Details |
| LiveCodeBench | Coding | 91.7% | Pass@1 | Unverified | Nov 18, 2025 | Details |
| METR Task-Completion Time Horizons | Agents | 224.3 min | 50% time horizon | Official | Nov 18, 2025 | Details |
| SciCode | Coding | 56.1% | accuracy | Unverified | Nov 18, 2025 | Details |
| MultiChallenge | Reasoning | 65.67% | accuracy | Verified | Nov 18, 2025 | Details |
| OpenAI-MRCR v2 (Multi-Round Coreference Resolution) | Reasoning | 77.0% | accuracy (mean SequenceMatcher similarity) | Verified | Nov 18, 2025 | Details |
| Global-MMLU | Reasoning | 92.2% | accuracy | Unverified | Nov 18, 2025 | Details |
| Video-MMMU | Multimodal | 87.6% | accuracy | Verified | Nov 18, 2025 | Details |
| WebDev Arena | Chat preference | 1438 | Elo | Verified | Nov 18, 2025 | Details |
| Search Arena | Chat preference | 1208 | Elo | Verified | Nov 18, 2025 | Details |
| EQ-Bench Creative Writing v3 | Chat preference | 1504 | Elo | Verified | Nov 18, 2025 | Details |
| Design Arena | Chat preference | 1295 | Elo | Verified | Nov 18, 2025 | Details |
| MASK (Model Alignment between Statements and Knowledge) | Other | 42.60 | Honesty score | Verified | Nov 18, 2025 | Details |
| MCP-Universe | Tool use | 44.59% | Overall Success Rate | Verified | Nov 18, 2025 | Details |
| CharXiv | Multimodal | 81.4% | accuracy | Unverified | Nov 18, 2025 | Details |
| OCRBench v2 | Multimodal | 63.4 | accuracy | Verified | Nov 18, 2025 | Details |
| ScreenSpot-Pro | Multimodal | 72.7% | accuracy | Verified | Nov 18, 2025 | Details |
| FACTS Grounding | Reasoning | 69.0% | Grounding accuracy | Verified | Nov 18, 2025 | Details |
| SWE-bench Multilingual | Coding | 68.7% | % resolved | Official | Nov 18, 2025 | Details |
| EnigmaEval | Reasoning | 18.24% | accuracy | Verified | Nov 18, 2025 | Details |
| ZeroBench | Multimodal | 19.0% (pass@5) | accuracy | Verified | Nov 18, 2025 | Details |
| MathArena HMMT February 2026 | Reasoning | 86.36% | accuracy | Official | Nov 18, 2025 | Details |
| FrontierMath Tier 4 | Reasoning | 18.8% | accuracy | Official | Nov 18, 2025 | Details |
| Vectara Hallucination Leaderboard | Other | 13.6% | Hallucination Rate | Official | Nov 18, 2025 | Details |
| Vibe Code Bench | Coding | 14.30% | Overall accuracy | Verified | Nov 18, 2025 | Details |
| GSO: Software Optimization Benchmark for SWE-Agents | Coding | 18.63% | Opt@1 | Official | Nov 18, 2025 | Details |
| MultiNRC | Reasoning | 58.96% | accuracy | Official | Nov 18, 2025 | Details |
| Remote Labor Index | Agents | 1.25% | automation rate | Official | Nov 18, 2025 | Details |
In the wild 1
See allReal-world feedback on Gemini 3 Pro from people using it on actual prompts — praise and criticism alike, each linked to its source. Qualitative, never scored.
Report tone
Report type
Topic
Mixed