Top 73
अब तक के सर्वश्रेष्ठ बेंचमार्क स्कोर
हर दर्ज मूल्यांकन में स्कोर के क्रम में।
Epoch Capabilities Index leads with 462, ahead of MMLU (217) and Terminal-Bench (203).
Benchmarks saturate: a test that separated models two years ago is often solved by everything today, which is why the set keeps growing and why an aggregate index exists at all. The Epoch Capabilities Index combines across them so progress stays visible after individual tests top out.
| # | Name | Metric | Latest model | Models scored |
|---|---|---|---|---|
| 01 | Epoch Capabilities Index | ECI Score | Jul 24, 2026 | 462 |
| 02 | MMLU | EM | Feb 26, 2025 | 217 |
| 03 | Terminal-Bench | Accuracy mean | Apr 23, 2026 | 203 |
| 04 | GPQA Diamond | mean_score | Jul 24, 2026 | 199 |
| 05 | OTIS Mock AIME 2024-2025 | mean_score | Jul 24, 2026 | 173 |
| 06 | GSM8K | EM | Nov 29, 2024 | 170 |
| 07 | ARC-AGI | Score | Jul 24, 2026 | 159 |
| 08 | ARC-AGI-2 | Score | Jul 24, 2026 | 159 |
| 09 | WeirdML | Accuracy | Jul 9, 2026 | 146 |
| 10 | WinoGrande | Accuracy | Dec 26, 2024 | 137 |
| 11 | BoolQ | Score | Aug 17, 2024 | 136 |
| 12 | ARC AI2 | Challenge score | Dec 26, 2024 | 134 |
| 13 | TriviaQA | EM | Dec 26, 2024 | 115 |
| 14 | PIQA | Score | Dec 26, 2024 | 113 |
| 15 | HellaSwag | Overall accuracy | Dec 26, 2024 | 113 |
| 16 | MATH Level 5 | mean_score | Oct 15, 2025 | 108 |
| 17 | FrontierMath | mean_score | May 28, 2026 | 101 |
| 18 | SimpleBench | Score (AVG@5) | Jul 9, 2026 | 91 |
| 19 | BIG-Bench Hard | Average | Dec 26, 2024 | 88 |
| 20 | ALE Bench | Performance | Jun 16, 2026 | 87 |
| 21 | Chess Puzzles | mean_score | Jul 24, 2026 | 85 |
| 22 | WebDev Arena | Arena Score | Jun 16, 2026 | 82 |
| 23 | SciCode | Score | Jul 24, 2026 | 75 |
| 24 | Critpt | Accuracy | Jul 24, 2026 | 74 |
| 25 | Aider Polyglot | Percent correct | Dec 1, 2025 | 72 |
| 26 | FrontierMath Tier 4 | mean_score | May 28, 2026 | 72 |
| 27 | OpenBookQA | Accuracy | Apr 23, 2024 | 70 |
| 28 | SimpleQA Verified | mean_score | Jul 24, 2026 | 68 |
| 29 | ForecastBench | Overall score | Mar 17, 2026 | 68 |
| 30 | LiveBench | Global average | Nov 13, 2025 | 54 |
| 31 | Vending-Bench 2 | Score | Jul 9, 2026 | 54 |
| 32 | LAMBADA | Score | Nov 30, 2023 | 53 |
| 33 | Humanity's Last Exam | Accuracy | Apr 16, 2026 | 51 |
| 34 | Apex Agents | Pass@1 score | Jul 24, 2026 | 50 |
| 35 | METR Time Horizons | average_score | Apr 7, 2026 | 50 |
| 36 | Video-MME | Overall (no subtitles) | Jun 18, 2025 | 50 |
| 37 | Lech Mazur Writing | Mean score | Aug 7, 2025 | 49 |
| 38 | Proofbench | Accuracy | Jul 24, 2026 | 44 |
| 39 | Enigma Eval | Accuracy | Mar 5, 2026 | 42 |
| 40 | Fictionlivebench | 120k token score | Jan 27, 2026 | 42 |
| 41 | Vpct | Correct | Dec 17, 2025 | 38 |
| 42 | Deepswe | Pass@1 | Jul 9, 2026 | 37 |
| 43 | Balrog | Average progress | Feb 19, 2026 | 36 |
| 44 | Deepresearchbench | Average score | May 28, 2026 | 36 |
| 45 | GSO | Score OPT@1 | Apr 23, 2026 | 36 |
| 46 | SWE-bench Verified | mean_score | Jun 16, 2026 | 35 |
| 47 | Geobench | ACW Avg Score | Dec 17, 2025 | 32 |
| 48 | Cursorbench | Score | Jul 24, 2026 | 31 |
| 49 | Posttrainbench | Average (%) | Jun 16, 2026 | 30 |
| 50 | ScienceQA | Score | Aug 16, 2024 | 26 |
| 51 | CL Bench | Overall | Mar 31, 2026 | 23 |
| 52 | Cybench | Unguided % Solved | Feb 5, 2026 | 22 |
| 53 | OSWorld | Score | Feb 17, 2026 | 20 |
| 54 | Exploitbench | Mean capability | Apr 23, 2026 | 19 |
| 55 | Algotune | Score | Mar 5, 2026 | 18 |
| 56 | CL Bench Life | Overall | Apr 24, 2026 | 17 |
| 57 | Blueprint Bench 2 | Score | Jul 9, 2026 | 17 |
| 58 | Surface Evolver Bench | Mean score | Jul 16, 2026 | 17 |
| 59 | Gbaeval | Overall score | Jun 16, 2026 | 16 |
| 60 | TheAgentCompany | % Score | Sep 29, 2025 | 16 |
| 61 | CAD Eval | Overall pass (%) | Apr 16, 2025 | 15 |
| 62 | Adversarial NLI | Score | Apr 23, 2024 | 15 |
| 63 | Frontiercode | Main score | Jul 24, 2026 | 14 |
| 64 | Frontierswe | Dominance | Jun 9, 2026 | 12 |
| 65 | GDPval | Win Rate (%) | Dec 11, 2025 | 11 |
| 66 | RLI | Score | Jun 9, 2026 | 10 |
| 67 | SuperGLUE | Score | Jan 27, 2022 | 10 |
| 68 | GDP PDF | GDP.pdf score | Jun 9, 2026 | 9 |
| 69 | OSWorld 2 | Binary accuracy | Jun 1, 2026 | 9 |
| 70 | Spatialviz Bench | Overall score | Jun 17, 2025 | 8 |
| 71 | Btf3 | Pooled score | Jul 9, 2026 | 8 |
| 72 | CommonsenseQA 2 | Score | Jul 18, 2023 | 6 |
| 73 | Mindcube | Overall score | May 22, 2025 | 5 |
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.
Browse all बेंचमार्क →