Compare models
Pick up to four models to see how they stack up on every benchmark they've all been scored on.
Scores from
Who comes out ahead?
Claude Fable 5.2 leads on 26 of the 62 benchmarks all four share. We use independent runs where they exist and fill the gaps with labs’ own numbers. 5 benchmarks are a tie for first.
- GPT-6.2 AstraBest on 13, tied for best on 4
- Claude Fable 5.2Best on 26, tied for best on 4
Gemini 4 CarbonBest on 17, tied for best on 2
Muse WonderBest on 1
Sources: Official leaderboards, Epoch AI (CC BY 4.0) and labs’ model cards
No shared benchmark matches these filters.
| AutomationBench | 52.4 | 51.5 | 53.9 | 39.1 (the lab’s own number) | 20 | 4 of 4 | 14.8 |
| FrontierCode 2.0 | 44.2 | 50.2 | 42.6 | 25.3 | 8 | 2 of 4 | 24.9 |
| LiveBench | 84.9 | 84.9 | 83.3 | 70.7 | 8 | 0 of 4 | 14.2 |
| LMArena (Elo)Elo | 1,468 | 1,503 | 1,527 | 1,511 | 7 | 0 of 4 | 59 |
| Agents' Last Exam v1.1 | 47.5 | 46.9 | 45.9 | 38.5 | 5 | 2 of 4 | 9.0 |
| Agents' Last Exam v1.1 (Score) | 78.3 | 72.8 | 77.4 | 68.4 | 5 | 2 of 4 | 10.0 |
| Artificial Analysis Intelligence Index v5index | 62.1 | 64.6 | 60.0 | 43.6 | 3 | 2 of 4 | 21.0 |
| HLE-Diamond | 72.3 | 77.9 | 63.6 | 40.6 | 3 | 0 of 4 | 37.3 |
| Legal Research Bench | 42.8 | 46.1 | 55.8 | 50.5 | 3 | 0 of 4 | 13.0 |
| Vals Index v2.2 | 67.7 | 71.3 | 70.5 | 64.9 | 2 | 1 of 4 | 6.4 |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 16.3 | 5.00 | 17.5 | 23.8 | 2 | 0 of 4 | 18.8 |
| Tax Agent Bench | 21.8 | 32.1 | 51.8 | 36.3 | 2 | 0 of 4 | 29.9 |
| Tax Agent Bench (Weighted pass) | 57.9 | 67.0 | 76.4 | 71.2 | 2 | 0 of 4 | 18.6 |
| Vibe Code Bench | 89.9 | 93.8 | 89.1 | 83.1 | 2 | 0 of 4 | 10.7 |
| APEX-Accounting | 69.8 | 70.4 | 70.7 | 67.1 | 1 | 0 of 4 | 3.6 |
| APEX-Accounting (Pass@1) | 19.5 | 22.5 | 22.5 | 21.3 | 1 | 0 of 4 | 3.0 |
| CWE-bench 2 | 66.8 | 75.0 | 67.3 | 66.1 | 1 | 0 of 4 | 8.9 |
| DesignArena Game DevElo | 1,450 | 1,392 | 1,329 | 1,409 | 1 | 0 of 4 | 121 |
| EEBench | 74.6 | 77.3 | 83.9 | 62.7 | 1 | 0 of 4 | 21.2 |
| Excel Modeling Benchmark v1.1 | 76.0 | 77.4 | 76.5 | 72.5 | 1 | 0 of 4 | 4.9 |
| FrontierMath Tier 4 v2 | 100.0 | 100.0 | 95.1 | 70.7 | 1 | 0 of 4 | 29.3 |
| FrontierMath Tiers 1-3 v2 | 92.6 | 95.1 | 90.5 | 82.1 | 1 | 0 of 4 | 13.0 |
| Humanity's Sixth Sense | 63.2 | 59.5 | 71.1 | 43.7 | 1 | 0 of 4 | 27.4 |
| Terminal-Bench 4.1 | 62.1 | 65.8 | 64.5 | 47.0 | 1 | 0 of 4 | 18.8 |
| Vals Finance Agent v2.1 | 60.8 | 69.5 | 67.2 | 65.5 | 1 | 0 of 4 | 8.7 |
| Vals Finance Index | 57.3 | 63.6 | 67.0 | 61.2 | 1 | 0 of 4 | 9.7 |
| Vals Legal Index | 69.7 | 69.2 | 75.9 | 75.3 | 1 | 0 of 4 | 6.8 |
| APEX-Agents 1.1 | 83.9 | 87.6 | 90.4 | 77.0 | 0 | 0 of 4 | 13.4 |
| Argo-Benchscore | 62.9 | 70.2 | 62.3 | 37.6 | 0 | 0 of 4 | 32.6 |
| BullshitBench V2 | 55.0 | 67.3 | 72.0 | 57.6 | 0 | 0 of 4 | 17.0 |
| BuseyBench SVG Testscore | 7.40 | 7.20 | 7.90 | 6.60 | 0 | 0 of 4 | 1.30 |
| CAIS Risk Indexscore, lower is better | 40.5 | 8.40 | 33.4 | 32.4 | 0 | 0 of 4 | 32.1 |
| CAIS Text Capabilities Indexscore | 54.7 | 66.3 | 59.0 | 47.6 | 0 | 0 of 4 | 18.7 |
| ContractorOS Quantity Takeoff | 53.8 | 59.2 | 60.0 | 31.5 | 0 | 0 of 4 | 28.5 |
| DataAnalyticsBenchscore | 41.0 | 43.0 | 42.0 | 41.0 | 0 | 0 of 4 | 2.00 |
| DAXBench v2 | 93.6 | 87.8 | 92.8 | 82.2 | 0 | 0 of 4 | 11.4 |
| DAYJOB: Accounting | 42.0 | 45.2 | 25.2 | 30.8 | 0 | 0 of 4 | 20.0 |
| DAYJOB: Law | 41.4 | 26.5 | 13.8 | 20.6 | 0 | 0 of 4 | 27.6 |
| DeepFinance | 61.4 | 62.0 | 56.6 | 44.2 | 0 | 0 of 4 | 17.8 |
| DesignArena PDFElo | 1,282 | 1,346 | 1,321 | 1,225 | 0 | 0 of 4 | 121 |
| DualEntry Accounting AI Benchmark | 78.2 | 81.2 | 88.1 | 83.2 | 0 | 0 of 4 | 9.9 |
| EBR-bench | 92.4 | 90.5 | 87.6 | 31.4 | 0 | 0 of 4 | 60.9 |
| Extended NYT Connections | 98.2 | 97.3 | 99.3 | 93.4 | 0 | 0 of 4 | 5.9 |
| Fiction Bench (Story Hacker)score | 73.6 | 62.1 | 64.4 | 64.0 | 0 | 0 of 4 | 11.5 |
| Furniture Assembly Benchmark | 89.2 | 90.0 | 77.5 | 54.2 | 0 | 0 of 4 | 35.8 |
| GradeBench | 56.3 | 57.3 | 56.6 | 48.1 | 0 | 0 of 4 | 9.2 |
| KradleBenchscore | 78.3 | 37.8 | 22.0 | 27.6 | 0 | 0 of 4 | 56.3 |
| Lech Mazur Creative Writingscore | 4.30 | 3.56 | 1.62 | 1.61 | 0 | 0 of 4 | 2.69 |
| LLM Debate Benchmarkrating | 1,673 | 1,688 | 1,676 | 1,637 | 0 | 0 of 4 | 51 |
| Maze Bench | 17.0 | 22.0 | 16.0 | 0.0 | 0 | 0 of 4 | 22.0 |
| Mystery Game Puzzles | 95.0 | 90.0 | 73.0 | 51.0 | 0 | 0 of 4 | 44.0 |
| OTIS Mock AIME 2024-2025 | 100.0 | 99.7 | 100.0 | 99.4 | 0 | 0 of 4 | 0.6 |
| OWE-Bench | 67.2 | 54.6 | 32.8 | 12.7 | 0 | 0 of 4 | 54.5 |
| Parallel Search Capability Leaderboardscore | 84.5 | 86.9 | 85.1 | 71.3 | 0 | 0 of 4 | 15.6 |
| Ramen Benchrating | 4 | 4 | 3 | 2 | 0 | 0 of 4 | 2 |
| Ramp Accounting Bench | 50.4 | 52.2 | 56.5 | 48.7 | 0 | 0 of 4 | 7.8 |
| Ramp SWE-Bench v2 | 76.4 | 73.4 | 74.0 | 58.6 | 0 | 0 of 4 | 17.8 |
| SWE Atlas v1.1 | 60.5 | 80.5 | 71.5 | 64.8 | 0 | 0 of 4 | 20.1 |
| TaxCalcBench v2 | 76.0 | 86.0 | 75.0 | 84.0 | 0 | 0 of 4 | 11.0 |
| The Aggregate Personal Style Imitation | 75.5 | 73.0 | 67.0 | 64.9 | 0 | 0 of 4 | 10.6 |
| Toloka Arena Enterprise Tool Use (November 2026) | 72.0 | 77.0 | 74.0 | 65.7 | 0 | 0 of 4 | 11.3 |
| Tuesday Work Indexindex | 70.3 | 71.2 | 65.6 | 63.8 | 0 | 0 of 4 | 7.4 |
A lab’s own number. Everything else is an independent run.Scores are percentages unless the row says otherwise. Bold is the best in each row. Click a row for its bars.
Sources: Official leaderboards, Epoch AI (CC BY 4.0) and labs’ model cards
Gap is how far apart the best and worst picked models are on a test. A small gap means they're about even. A big one means the test tells them apart.
It's in points for percentages (74.3 against 68.1 is 6.2) and in the test's own unit otherwise, such as Elo. To sort mixed units together, we count those as a share of the test's full range.
Labs is how many labs report this benchmark in their launch posts and system cards, across the whole almanac. A high count means a widely used test.
Cited by is how many of the picked models' own launch posts and system cards report this benchmark. At 0, every score in the row is an independent run.