Every evaluation tracked, from saturated classics to the frontier tests models still fail — plus the Epoch Capabilities Index that ties them together.
| Epoch Capabilities Index | ECI Score | 462 | 161.7 | Jul 24, 2026 |
| MMLU | EM | 217 | 0.88 | Feb 26, 2025 |
| Terminal-Bench | Accuracy mean | 203 | 0.9 | Apr 23, 2026 |
| GPQA Diamond | mean_score | 199 | 0.95 | Jul 24, 2026 |
| OTIS Mock AIME 2024-2025 | mean_score | 173 | 1 | Jul 24, 2026 |
| GSM8K | EM | 170 | 0.94 | Nov 29, 2024 |
| ARC-AGI-2 | Score | 159 | 0.93 | Jul 24, 2026 |
| ARC-AGI | Score | 159 | 0.98 | Jul 24, 2026 |
| WeirdML | Accuracy | 146 | 0.89 | Jul 9, 2026 |
| WinoGrande | Accuracy | 137 | 0.89 | Dec 26, 2024 |
| BoolQ | Score | 136 | 0.91 | Aug 17, 2024 |
| ARC AI2 | Challenge score | 134 | 0.95 | Dec 26, 2024 |
| TriviaQA | EM | 115 | 0.88 | Dec 26, 2024 |
| HellaSwag | Overall accuracy | 113 | 0.95 | Dec 26, 2024 |
| PIQA | Score | 113 | 0.89 | Dec 26, 2024 |
| MATH Level 5 | mean_score | 108 | 0.98 | Oct 15, 2025 |
| FrontierMath | mean_score | 101 | 0.52 | May 28, 2026 |
| SimpleBench | Score (AVG@5) | 91 | 0.82 | Jul 9, 2026 |
| BIG-Bench Hard | Average | 88 | 0.89 | Dec 26, 2024 |
| ALE Bench | Performance | 87 | 2041.3 | Jun 16, 2026 |
| Chess Puzzles | mean_score | 85 | 0.64 | Jul 24, 2026 |
| WebDev Arena | Arena Score | 82 | 1653.9 | Jun 16, 2026 |
| SciCode | Score | 75 | 0.6 | Jul 24, 2026 |
| Critpt | Accuracy | 74 | 0.32 | Jul 24, 2026 |
| Aider Polyglot | Percent correct | 72 | 88 | Dec 1, 2025 |
| FrontierMath Tier 4 | mean_score | 72 | 0.48 | May 28, 2026 |
| OpenBookQA | Accuracy | 70 | 0.88 | Apr 23, 2024 |
| SimpleQA Verified | mean_score | 68 | 0.77 | Jul 24, 2026 |
| ForecastBench | Overall score | 68 | 62.5 | Mar 17, 2026 |
| LiveBench | Global average | 54 | 82.3 | Nov 13, 2025 |
| Vending-Bench 2 | Score | 54 | 10936.8 | Jul 9, 2026 |
| LAMBADA | Score | 53 | 0.87 | Nov 30, 2023 |
| Humanity's Last Exam | Accuracy | 51 | 0.46 | Apr 16, 2026 |
| Video-MME | Overall (no subtitles) | 50 | 0.8 | Jun 18, 2025 |
| Apex Agents | Pass@1 score | 50 | 0.5 | Jul 24, 2026 |
| METR Time Horizons | average_score | 50 | 0.85 | Apr 7, 2026 |
| Lech Mazur Writing | Mean score | 49 | 8.6 | Aug 7, 2025 |
| Proofbench | Accuracy | 44 | 0.78 | Jul 24, 2026 |
| Enigma Eval | Accuracy | 42 | 0.24 | Mar 5, 2026 |
| Fictionlivebench | 120k token score | 42 | 1 | Jan 27, 2026 |
| Vpct | Correct | 38 | 0.91 | Dec 17, 2025 |
| Deepswe | Pass@1 | 37 | 0.73 | Jul 9, 2026 |
| Deepresearchbench | Average score | 36 | 0.55 | May 28, 2026 |
| Balrog | Average progress | 36 | 0.58 | Feb 19, 2026 |
| GSO | Score OPT@1 | 36 | 0.44 | Apr 23, 2026 |
| SWE-bench Verified | mean_score | 35 | 0.83 | Jun 16, 2026 |
| Geobench | ACW Avg Score | 32 | 4333 | Dec 17, 2025 |
| Cursorbench | Score | 31 | 0.73 | Jul 24, 2026 |
| Posttrainbench | Average (%) | 30 | 0.34 | Jun 16, 2026 |
| ScienceQA | Score | 26 | 0.91 | Aug 16, 2024 |
| CL Bench | Overall | 23 | 0.28 | Mar 31, 2026 |
| Cybench | Unguided % Solved | 22 | 0.93 | Feb 5, 2026 |
| OSWorld | Score | 20 | 72.1 | Feb 17, 2026 |
| Exploitbench | Mean capability | 19 | 0.69 | Apr 23, 2026 |
| Algotune | Score | 18 | 2.05 | Mar 5, 2026 |
| Surface Evolver Bench | Mean score | 17 | 0.95 | Jul 16, 2026 |
| CL Bench Life | Overall | 17 | 0.22 | Apr 24, 2026 |
| Blueprint Bench 2 | Score | 17 | 0.39 | Jul 9, 2026 |
| TheAgentCompany | % Score | 16 | 0.52 | Sep 29, 2025 |
| Gbaeval | Overall score | 16 | 0.74 | Jun 16, 2026 |
| CAD Eval | Overall pass (%) | 15 | 0.74 | Apr 16, 2025 |
| Adversarial NLI | Score | 15 | 0.58 | Apr 23, 2024 |
| Frontiercode | Main score | 14 | 0.54 | Jul 24, 2026 |
| Frontierswe | Dominance | 12 | 0.9 | Jun 9, 2026 |
| GDPval | Win Rate (%) | 11 | 0.5 | Dec 11, 2025 |
| SuperGLUE | Score | 10 | 0.89 | Jan 27, 2022 |
| RLI | Score | 10 | 0.16 | Jun 9, 2026 |
| OSWorld 2 | Binary accuracy | 9 | 0.21 | Jun 1, 2026 |
| GDP PDF | GDP.pdf score | 9 | 0.3 | Jun 9, 2026 |
| Spatialviz Bench | Overall score | 8 | 0.45 | Jun 17, 2025 |
| Btf3 | Pooled score | 8 | 0.15 | Jul 9, 2026 |
| CommonsenseQA 2 | Score | 6 | 0.68 | Jul 18, 2023 |
| Mindcube | Overall score | 5 | 0.47 | May 22, 2025 |
