Raw scores stay raw
No price- or context-derived performance scores.
Compare like with like.
Raw results remain attached to their benchmark version, metric, source, and evaluation date. Pick any row to rank compatible models.
Version-aware
Incompatible benchmark versions are never silently mixed.
Evidence labeled
Source-linked and independently cross-checked results are visibly distinct.
| Benchmark | Version | Metric | Models | Sources | Latest evaluation | |
|---|---|---|---|---|---|---|
| Design Arena: website | Unversioned | elo | 106 | 1 | Not documented | Rank models → |
| Coding Index | Unversioned | index | 105 | 1 | Not documented | Rank models → |
| Design Arena: gamedev | Unversioned | elo | 96 | 1 | Not documented | Rank models → |
| Agentic Index | Unversioned | index | 96 | 1 | Not documented | Rank models → |
| Design Arena: codecategories | Unversioned | elo | 96 | 1 | Not documented | Rank models → |
| Design Arena: dataviz | Unversioned | elo | 95 | 1 | Not documented | Rank models → |
| Intelligence Index | Unversioned | index | 95 | 1 | Not documented | Rank models → |
| Design Arena: uicomponent | Unversioned | elo | 90 | 1 | Not documented | Rank models → |
| Design Arena: 3d | Unversioned | elo | 87 | 1 | Not documented | Rank models → |
| Design Arena: svg | Unversioned | elo | 66 | 1 | Not documented | Rank models → |
| Design Arena: asciiart | Unversioned | elo | 46 | 1 | Not documented | Rank models → |
| SWE-Bench Pro | Unversioned | resolve rate | 38 | 17 | 2026-07-09 | Rank models → |
| SWE-Bench Verified | Unversioned | resolved | 34 | 28 | 2026-07-02 | Rank models → |
| Design Arena: fullstack | Unversioned | elo | 31 | 1 | Not documented | Rank models → |
| Aider Polyglot | Unversioned | percent correct | 31 | 1 | 2025-10-03 | Rank models → |
| Design Arena: mobileapps | Unversioned | elo | 30 | 1 | Not documented | Rank models → |
| Design Arena: webapps | Unversioned | elo | 28 | 1 | Not documented | Rank models → |
| Design Arena: androidnative | Unversioned | elo | 26 | 1 | Not documented | Rank models → |
| SciCode | Unversioned | percent correct | 26 | 25 | 2026-06-04 | Rank models → |
| Artificial Analysis Coding Index | Unversioned | index | 25 | 24 | 2026-06-04 | Rank models → |
| Terminal-Bench Hard | Unversioned | success rate | 21 | 20 | 2026-06-04 | Rank models → |
| Terminal-Bench | 2.1 | success rate | 19 | 12 | 2026-07-09 | Rank models → |
| Design Arena: godotgamedev | Unversioned | elo | 18 | 1 | Not documented | Rank models → |
| Humanity's Last Exam | Unversioned | accuracy | 14 | 8 | 2026-07-09 | Rank models → |
| Design Arena: htmlslides | Unversioned | elo | 13 | 1 | Not documented | Rank models → |
| Design Arena: agenticgamedev | Unversioned | elo | 12 | 1 | Not documented | Rank models → |
| BrowseComp | Unversioned | accuracy | 12 | 7 | 2026-07-09 | Rank models → |
| GPQA Diamond | Unversioned | accuracy | 12 | 6 | 2026-07-09 | Rank models → |
| Design Arena: python-pptxslides | Unversioned | elo | 11 | 1 | Not documented | Rank models → |
| OSWorld-Verified | Unversioned | success rate | 11 | 6 | 2026-07-09 | Rank models → |
| SWE-Atlas Refactoring | Unversioned | score | 11 | 1 | Not documented | Rank models → |
| SWE-Atlas Codebase QnA | Unversioned | score | 10 | 1 | Not documented | Rank models → |
| Design Arena: agenticslides(python-pptx) | Unversioned | elo | 10 | 1 | Not documented | Rank models → |
| Design Arena: agenticslides(html) | Unversioned | elo | 10 | 1 | Not documented | Rank models → |
| Design Arena: agenticslides | Unversioned | elo | 10 | 1 | Not documented | Rank models → |
| Design Arena: agentichtmlslides | Unversioned | elo | 10 | 1 | Not documented | Rank models → |
| SWE-Atlas Test Writing | Unversioned | score | 10 | 1 | Not documented | Rank models → |
| Terminal-Bench | 2.1 | pass@1 | 9 | 1 | Not documented | Rank models → |
| SWE-Bench Pro | Unversioned | pass@1 | 9 | 1 | Not documented | Rank models → |
| Artificial Analysis Coding Agent Index | Unversioned | average pass@1 | 9 | 1 | Not documented | Rank models → |
| SWE-Atlas Codebase QnA | Unversioned | pass@1 | 9 | 1 | Not documented | Rank models → |
| Design Arena: pptxslides | Unversioned | elo | 9 | 1 | Not documented | Rank models → |
| MCP Atlas | Unversioned | success rate | 7 | 6 | 2026-07-09 | Rank models → |
| MMMU Pro | Unversioned | accuracy | 7 | 3 | 2026-07-09 | Rank models → |
| GDPval | Unversioned | wins or ties | 6 | 3 | 2026-06-04 | Rank models → |
| SWE-Bench Multilingual | Unversioned | resolve rate | 6 | 6 | 2026-07-08 | Rank models → |
| Toolathlon | Unversioned | success rate | 6 | 4 | 2026-07-09 | Rank models → |
| Design Arena: logo | Unversioned | elo | 5 | 1 | Not documented | Rank models → |
| DeepSWE | 1.1 | resolve rate | 5 | 3 | 2026-07-09 | Rank models → |
| Design Arena: graphicdesign | Unversioned | elo | 5 | 1 | Not documented | Rank models → |
| ARC-AGI-2 | Unversioned | accuracy | 5 | 2 | 2026-05-19 | Rank models → |
| Terminal-Bench | 2.0 | success rate | 5 | 4 | 2026-07-02 | Rank models → |
| Design Arena: image | Unversioned | elo | 5 | 1 | Not documented | Rank models → |
| GDPval-AA | Unversioned | Elo | 4 | 3 | 2026-06-09 | Rank models → |
| FrontierMath | Unversioned | accuracy | 4 | 1 | 2026-04-23 | Rank models → |
| SWE-Bench Pro | Unversioned | percent resolved | 3 | 3 | Not documented | Rank models → |
| FrontierMath | v2 | accuracy | 3 | 1 | 2026-07-09 | Rank models → |
| Claw-eval | Unversioned | percent | 3 | 3 | Not documented | Rank models → |
| NL2Repo | Unversioned | percent | 3 | 3 | Not documented | Rank models → |
| OSWorld | 2.0 | success rate | 3 | 1 | 2026-07-09 | Rank models → |
| CharXiv Reasoning | Unversioned | accuracy | 3 | 2 | 2026-07-09 | Rank models → |
| SWE-Bench Multilingual | Unversioned | percent resolved | 3 | 3 | Not documented | Rank models → |
| Artificial Analysis Intelligence Index | 4.1 | index score | 3 | 1 | 2026-07-09 | Rank models → |
| SWE-Bench Verified | Unversioned | percent resolved | 3 | 3 | Not documented | Rank models → |
| SciCode | Unversioned | score | 3 | 2 | 2026-05-19 | Rank models → |
| Artificial Analysis Coding Agent Index | 1.1 | index score | 3 | 1 | 2026-07-09 | Rank models → |
| Agents' Last Exam | Unversioned | score | 3 | 1 | 2026-07-09 | Rank models → |
| Terminal-Bench 2.1 | Unversioned | percent | 3 | 3 | Not documented | Rank models → |
| τ²-Bench Telecom | Unversioned | success rate | 3 | 3 | 2026-06-09 | Rank models → |
| MRCRv2 | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| CTI-REALM | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| Long Context Reasoning | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| LiveCodeBench | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| Terminal Bench 2.1 | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| τ3 Banking | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| GPQA Diamond | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| GeneBench | Unversioned | accuracy | 2 | 1 | 2026-04-23 | Rank models → |
| CharXiv Reasoning | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| Humanity’s Last Exam | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| IFBench | Unversioned | accuracy | 2 | 2 | 2026-06-04 | Rank models → |
| FrontierCode | Unversioned | pass rate | 2 | 1 | 2026-06-09 | Rank models → |
| LiveCodeBench Pro | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| SWE Bench Pro | Unversioned | score | 2 | 1 | Not documented | Rank models → |
| AutomationBench | Unversioned | success rate | 1 | 1 | 2026-06-09 | Rank models → |
| Design Arena: imageediting | Unversioned | elo | 1 | 1 | Not documented | Rank models → |
| Toolathlon-Verified | Unversioned | success rate | 1 | 1 | 2026-07-09 | Rank models → |
| Artificial Analysis Intelligence Index | 4.0 | index score | 1 | 1 | 2026-04-30 | Rank models → |
| FORTE | Unversioned | success rate | 1 | 1 | 2026-06-30 | Rank models → |
| Artificial Analysis Intelligence Index | Unversioned | index score | 1 | 1 | 2026-06-09 | Rank models → |
| Artificial Analysis Coding Index | Unversioned | index score | 1 | 1 | 2026-06-09 | Rank models → |
| ARC-AGI-1 | Unversioned | accuracy | 1 | 1 | 2026-04-23 | Rank models → |
| GDPval-AA | Unversioned | win rate | 1 | 1 | 2026-06-09 | Rank models → |
| GPQA | Unversioned | accuracy | 1 | 1 | 2026-06-04 | Rank models → |
| FrontierSWE | Unversioned | dominance | 1 | 1 | 2026-06-16 | Rank models → |
| FrontierCode | v1 | pass rate | 1 | 1 | 2026-06-30 | Rank models → |
| IFEval | Unversioned | accuracy | 1 | 1 | 2026-06-30 | Rank models → |
| DeepSWE | 1.0 | resolve rate | 1 | 1 | 2026-07-08 | Rank models → |
| Kimi Claw 24/7 Bench | Unversioned | score | 1 | 1 | 2026-06-12 | Rank models → |
| JobBench | Unversioned | success rate | 1 | 1 | 2026-07-09 | Rank models → |
| Kimi Code Bench | v2 | score | 1 | 1 | 2026-06-12 | Rank models → |
| LMArena Text-to-Video Arena | Unversioned | Elo | 1 | 1 | 2026-06-30 | Rank models → |
| LiveCodeBench | v6 | pass@1 | 1 | 1 | 2026-06-04 | Rank models → |
| ClawEval | 1.1 | pass^3 | 1 | 1 | 2026-05-29 | Rank models → |
| MCP Mark Verified | Unversioned | success rate | 1 | 1 | 2026-06-12 | Rank models → |
| MLS Bench Lite | Unversioned | score | 1 | 1 | 2026-06-12 | Rank models → |
| MMLU-Pro | Unversioned | accuracy | 1 | 1 | 2026-06-04 | Rank models → |
| Finance Agent | v2 | accuracy | 1 | 1 | 2026-07-09 | Rank models → |
| Terminal Bench 2.0 | Unversioned | score | 1 | 1 | Not documented | Rank models → |
| NL2Repo | Unversioned | score | 1 | 1 | 2026-05-19 | Rank models → |
| Program Bench | Unversioned | score | 1 | 1 | 2026-06-12 | Rank models → |
| FinanceAgent | 1.1 | accuracy | 1 | 1 | 2026-04-23 | Rank models → |
| BabyVision | Unversioned | accuracy | 1 | 1 | 2026-07-09 | Rank models → |