OpenAIo-series
o3
OpenAI · o-series. Released Apr 16, 2025.
o3 is a model from OpenAI in the o-series family, released Apr 16, 2025. evals.report tracks 44 reported o3 benchmark scores across FrontierMath, Berkeley Function Calling Leaderboard, ARC-AGI-1, ARC-AGI-2, SWE-bench Verified, MMMU-Pro, SimpleQA Verified, GPQA Diamond, and 36 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.
44 results
Benchmark results 44
Compare this model| Benchmark | Category | Score | Metric | Status | Date | |
|---|---|---|---|---|---|---|
| FrontierMath | Reasoning | 18.69% | accuracy | Official | Apr 16, 2025 | Details |
| Berkeley Function Calling Leaderboard | Tool use | 63.05% | accuracy | Official | Apr 16, 2025 | Details |
| ARC-AGI-1 | Reasoning | 60.83% | accuracy | Official | Apr 16, 2025 | Details |
| ARC-AGI-2 | Reasoning | 6.53% | accuracy | Official | Apr 16, 2025 | Details |
| SWE-bench Verified | Coding | 62.3% | % resolved | Official | Apr 16, 2025 | Details |
| MMMU-Pro | Multimodal | 76.4% | accuracy | Official | Apr 16, 2025 | Details |
| SimpleQA Verified | Other | 53.0% | accuracy | Official | Apr 16, 2025 | Details |
| GPQA Diamond | Reasoning | 81.8% | accuracy | Official | Apr 16, 2025 | Details |
| AIME (OTIS Mock) | Reasoning | 83.9% | accuracy | Official | Apr 16, 2025 | Details |
| LiveCodeBench Pro | Coding | 1010 | Codeforces Elo | Official | Apr 16, 2025 | Details |
| WeirdML | Coding | 52.4% | average accuracy | Official | Apr 16, 2025 | Details |
| Artificial Analysis Intelligence Index | Reasoning | 38.4 | Index | Unverified | Apr 16, 2025 | Details |
| Epoch Capabilities Index | Reasoning | 147.3 | Index | Official | Apr 16, 2025 | Details |
| Aider Polyglot | Coding | 81.3% | % correct | Official | Apr 16, 2025 | Details |
| MMLU-Pro | Reasoning | 85.3% | accuracy | Verified | Apr 16, 2025 | Details |
| GAIA: A Benchmark for General AI Assistants | Agents | 32.73% | accuracy | Unverified | Apr 16, 2025 | Details |
| BrowseComp | Agents | 49.7% | accuracy | Verified | Apr 16, 2025 | Details |
| τ²-bench (Telecom) | Tool use | 58.2% | pass^1 | Unverified | Apr 16, 2025 | Details |
| MathVista | Multimodal | 86.8% | accuracy | Verified | Apr 16, 2025 | Details |
| GDPval | Agents | 753 | Elo | Official | Apr 16, 2025 | Details |
| LiveCodeBench | Coding | 80.8% | Pass@1 | Unverified | Apr 16, 2025 | Details |
| METR Task-Completion Time Horizons | Agents | 119.7 min | 50% time horizon | Official | Apr 16, 2025 | Details |
| SciCode | Coding | 41.0% | accuracy | Unverified | Apr 16, 2025 | Details |
| MMMU (Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark) | Multimodal | 82.9% | accuracy | Verified | Apr 16, 2025 | Details |
| IFBench | Reasoning | 69.3% | accuracy | Official | Apr 16, 2025 | Details |
| MultiChallenge | Reasoning | 56.62% | accuracy | Verified | Apr 16, 2025 | Details |
| Video-MMMU | Multimodal | 83.3% | accuracy | Unverified | Apr 16, 2025 | Details |
| Search Arena | Chat preference | 1144 | Elo | Verified | Apr 16, 2025 | Details |
| Arena-Hard-Auto v2.0 | Chat preference | 85.9% | % win rate | Official | Apr 16, 2025 | Details |
| EQ-Bench Creative Writing v3 | Chat preference | 1744 | Elo | Verified | Apr 16, 2025 | Details |
| Design Arena | Chat preference | 1074 | Elo | Verified | Apr 16, 2025 | Details |
| MASK (Model Alignment between Statements and Knowledge) | Other | 84.47 | Honesty score | Verified | Apr 16, 2025 | Details |
| MCP-Universe | Tool use | 26.41% | Overall Success Rate | Verified | Apr 16, 2025 | Details |
| CharXiv | Multimodal | 78.6% | accuracy | Unverified | Apr 16, 2025 | Details |
| FACTS Grounding | Reasoning | 36.2% | Grounding accuracy | Verified | Apr 16, 2025 | Details |
| SWE-bench Multimodal | Coding | 35.98% | % resolved | Verified | Apr 16, 2025 | Details |
| EnigmaEval | Reasoning | 13.09% | accuracy | Verified | Apr 16, 2025 | Details |
| ZeroBench | Multimodal | 3.0% (pass@1) | accuracy | Verified | Apr 16, 2025 | Details |
| IMO-Bench | Reasoning | 61.1% | accuracy | Verified | Apr 16, 2025 | Details |
| FrontierMath Tier 4 | Reasoning | 2.1% | accuracy | Official | Apr 16, 2025 | Details |
| Gray Swan Arena (Agent Red-Teaming / Indirect Prompt Injection) | Agents | 2.50% | Attack Success Rate (ASR) | Verified | Apr 16, 2025 | Details |
| Online-Mind2Web | Agents | 39.00% | Task success rate | Verified | Apr 16, 2025 | Details |
| GSO: Software Optimization Benchmark for SWE-Agents | Coding | 8.82% | Opt@1 | Official | Apr 16, 2025 | Details |
| MultiNRC | Reasoning | 45.50% | accuracy | Official | Apr 16, 2025 | Details |