190 public LLM benchmarks
- Coding
- 38
- Intelligence
- 2
- Agentic work
- 39
- Reasoning
- 21
190 public boards, by category · read October 20
- Coding: 38 boards
- Intelligence: 2 boards
- Agentic work: 39 boards
- Reasoning: 21 boards
- Math: 13 boards
- Knowledge: 17 boards
Every board we read, grouped by what it tests, as of October 20, 2026. Each one links to its own page with the full ranking, what it measures and the source it was read from. We publish results as each board does and never fill in a missing one.
Coding
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| LMArena Text: coding | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1547 Elo |
| LMArena Text: software and IT | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1556 Elo |
| Terminal-Bench Hard (Artificial Analysis run) | Artificial Analysis | 126 | GPT-5.6 Sol65.9% |
| AA Coding Index | Artificial Analysis (via OpenRouter) | 114 | Claude Fable 5.181.6 |
| SciCode | Artificial Analysis | 108 | Claude Opus 5.566.9% |
| ALE-Bench | ALE-Bench (Sakana AI) | 104 | GPT-6 Astra2951.3 |
| LMArena WebDevIn the coding index | LMArena WebDev | 101 | Claude Opus 5.51807 Elo |
| LMArena WebDev: HTML | LMArena (Hugging Face dataset) | 100 | GPT-6.1 Sol1800 Elo |
| Vibe Code Bench (Vals AI) | Vals AI boards | 96 | Claude Sonnet 5.592.4% |
| LiveCodeBench (Vals AI) | Vals AI boards | 95 | Claude Fable 5.190.5% |
| WeirdML | Epoch AI benchmarking hub | 93 | GPT-6 Astra Pro93.6% |
| LMArena WebDev: React | LMArena (Hugging Face dataset) | 92 | Claude Opus 5.51855 Elo |
| LiveCodeBench (Artificial Analysis run) | Artificial Analysis | 84 | Gemini 3 Flash Preview90.8% |
| Code migration (Vals AI) | Vals AI boards | 76 | Claude Sonnet 5.569.8% |
| SWE-bench Verified (Vals AI) | Vals AI boards | 76 | Claude Opus 597.0% |
| Terminal-Bench 2.1 (Vals AI) | Vals AI boards | 68 | Claude Opus 5.587.6% |
| ProgramBench (Vals AI) | Vals AI boards | 59 | Claude Opus 5.518.5% |
| LMArena WebDev: from an image | LMArena (Hugging Face dataset) | 49 | Claude Opus 5.51753 Elo |
| Terminal-Bench 4 (Vals AI) | Vals AI boards | 48 | Claude Opus 5.565.2% |
| LiveBench codingIn the coding index | LiveBench | 43 | Claude Fable 598.3% |
| IOI (Vals AI) | Vals AI boards | 42 | Gemini 4 Argon100.0% |
| CyberBench (Vals AI) | Vals AI boards | 41 | Muse Wonder80.4% |
| FrontierCode | Epoch AI benchmarking hub | 41 | Claude Opus 5.554.6% |
| Terminal-Bench | Epoch AI benchmarking hub | 40 | GPT-5.584.7% |
| SWE-bench VerifiedIn the coding index | SWE-bench (swebench.com) | 38 | Claude Opus 4.579.2% |
| SkillsBench (Vals AI) | Vals AI boards | 34 | DeepSeek V4.1 Flash69.8% |
| DeepSWE | Epoch AI benchmarking hub | 26 | GPT-6 Astra74.1% |
| GSO-Bench | Epoch AI benchmarking hub | 26 | Claude Fable 5.188.2% |
| Vibe Code Bench 1-100 (Vals AI) | Vals AI boards | 23 | Claude Opus 5.530.4% |
| Aider polyglotIn the coding index | Aider polyglot leaderboard | 22 | GPT-588.0% |
| FrontierSWE | Epoch AI benchmarking hub | 20 | GPT-6 Astra65.5% |
| SWE-bench Multimodal (Vals AI) | Vals AI boards | 18 | Claude Opus 5.570.2% |
| SWE-bench Multilingual | SWE-bench (swebench.com) | 11 | Claude Opus 4.672.0% |
| MirrorCode | Epoch AI benchmarking hub | 9 | Claude Opus 5.577.4% |
| Cybench | Epoch AI benchmarking hub | 6 | Claude Opus 4.693.0% |
| SWE-bench Lite | SWE-bench (swebench.com) | 6 | Qwen3 Coder 30B A3B Instruct49.7% |
| SWE-bench Multimodal | SWE-bench (swebench.com) | 4 | o336.0% |
| SWE-bench (full test set) | SWE-bench (swebench.com) | 3 | Claude Opus 4.552.6% |
Intelligence
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| AA Intelligence Index | Artificial Analysis (via OpenRouter) | 213 | Claude Opus 5.557.6 |
| Epoch Capabilities Index | Epoch AI benchmarking hub | 132 | Claude Opus 5.5166.8 |
Agentic work
Reasoning
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| Humanity's Last Exam (Artificial Analysis run) | Artificial Analysis | 192 | Claude Opus 5.561.4% |
| GPQA Diamond (Artificial Analysis run) | Artificial Analysis | 175 | GPT-6 Astra96.3% |
| LMArena Text: hard prompts | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1546 Elo |
| LMArena Text: hard prompts in English | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1548 Elo |
| LMArena Text: expert | LMArena (Hugging Face dataset) | 138 | Claude Opus 51555 Elo |
| GPQA Diamond (Epoch AI run) | Epoch AI benchmarking hub | 116 | Gemini 4 Argon96.4% |
| GPQA Diamond | GPQA (Vals AI) | 91 | Gemini 3.1 Pro Preview95.5% |
| Chess puzzles | Epoch AI benchmarking hub | 81 | GPT-6 Astra Pro76.0% |
| LMArena Vision: diagrams | LMArena (Hugging Face dataset) | 76 | Claude Fable 51357 Elo |
| LMArena Vision: homework | LMArena (Hugging Face dataset) | 74 | Claude Opus 5.51362 Elo |
| ARC-AGI | Epoch AI benchmarking hub | 73 | Claude Fable 598.5% |
| Mystery game puzzles | Epoch AI benchmarking hub | 73 | GPT-6 Astra84.0% |
| ARC-AGI-2 | Epoch AI benchmarking hub | 72 | GPT-6 Astra95.0% |
| MMMU Pro (Vals AI) | Vals AI boards | 63 | Claude Fable 5.190.6% |
| SimpleBench | Epoch AI benchmarking hub | 56 | Claude Fable 581.9% |
| LiveBench reasoning | LiveBench | 43 | Gemini 4 Argon95.1% |
| Humanity's Last Exam | Epoch AI benchmarking hub | 34 | GPT-6 Astra54.8% |
| EBR-bench | Epoch AI benchmarking hub | 26 | GPT-6 Astra76.2% |
| MysteryMechanism (Vals AI) | Vals AI boards | 25 | GPT-6 Astra53.1% |
| VPCT | Epoch AI benchmarking hub | 15 | GPT-5.284.0% |
| BIG-Bench Hard | Epoch AI benchmarking hub | 3 | Qwen2.5 72B Instruct79.8% |
Math
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| LMArena Text: math | LMArena (Hugging Face dataset) | 135 | Gemini 4 Argon1536 Elo |
| LMArena Text: mathematical work | LMArena (Hugging Face dataset) | 135 | Gemini 4 Argon1556 Elo |
| OTIS Mock AIME | Epoch AI benchmarking hub | 116 | Claude Fable 5100.0% |
| FrontierMath, tiers 1 to 3 | Epoch AI benchmarking hub | 79 | GPT-6 Astra93.7% |
| ProofBench | Epoch AI benchmarking hub | 74 | Claude Fable 5.1100.0% |
| AA Math Index | Artificial Analysis | 72 | GPT-5.299.0 |
| AIME 2025 (Artificial Analysis run) | Artificial Analysis | 72 | GPT-5.299.0% |
| FrontierMath, tier 4 | Epoch AI benchmarking hub | 64 | GPT-6.1 Sol100.0% |
| AIME | AIME (Vals AI) | 54 | Gemini 3.1 Pro Preview98.1% |
| MGSM (Vals AI) | Vals AI boards | 37 | Claude Opus 4.595.2% |
| MATH level 5 | Epoch AI benchmarking hub | 28 | GPT-598.1% |
| MATH 500 (Vals AI) | Vals AI boards | 25 | GPT-596.0% |
| GSM8K | Epoch AI benchmarking hub | 6 | Qwen2.5 Coder 32B Instruct93.0% |
Knowledge
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| LMArena Text: business and finance | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1525 Elo |
| LMArena Text: medicine and healthcare | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1554 Elo |
| LMArena Text: science | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1548 Elo |
| LMArena Text: legal and government | LMArena (Hugging Face dataset) | 138 | Gemini 4 Argon1549 Elo |
| LegalBench (Vals AI) | Vals AI boards | 95 | Claude Fable 588.6% |
| MMLU Pro | MMLU Pro (Vals AI) | 92 | Claude Fable 5.192.4% |
| TaxEval (Vals AI) | Vals AI boards | 92 | Muse Spark 1.280.4% |
| CorpFin (Vals AI) | Vals AI boards | 86 | Claude Opus 573.2% |
| MMLU Pro (Artificial Analysis run) | Artificial Analysis | 86 | Claude Opus 4.589.5% |
| MedCode (Vals AI) | Vals AI boards | 83 | Claude Opus 563.6% |
| SimpleQA Verified | Epoch AI benchmarking hub | 76 | Gemini 4 Argon76.0% |
| MortgageTax (Vals AI) | Vals AI boards | 68 | Claude Opus 572.1% |
| MedQA (Vals AI) | Vals AI boards | 47 | o196.5% |
| CaseLaw (Vals AI) | Vals AI boards | 45 | Grok 4.379.3% |
| MMLU | Epoch AI benchmarking hub | 23 | GPT-4o88.1% |
| GeoBench | Epoch AI benchmarking hub | 12 | Gemini 3 Flash Preview88.0% |
| AI2 Reasoning Challenge | Epoch AI benchmarking hub | 3 | GPT-496.3% |
Web design
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| Design Arena: websites | Design Arena | 131 | Muse Wonder1360 Elo |
| Design Arena: game dev | Design Arena | 127 | GPT-6 Astra1474 Elo |
| Design Arena: all code categories | Design Arena | 125 | Claude Opus 5.51390 Elo |
| Design Arena: UI components | Design Arena | 120 | GPT-6 Astra1400 Elo |
| Design Arena: data viz | Design Arena | 119 | Claude Opus 5.51397 Elo |
| Design Arena: 3D | Design Arena | 116 | GPT-6 Astra1505 Elo |
| Design Arena: SVG | Design Arena | 87 | GPT-6 Astra1431 Elo |
| Design Arena: ASCII art | Design Arena | 73 | GPT-6 Astra1356 Elo |
| LiveBench design | LiveBench | 43 | Claude Opus 5.586.9% |
General
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| Long context reasoning (AA-LCR) | Artificial Analysis | 178 | Kimi K388.7% |
| LMArena Text | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1532 Elo |
| LMArena Text: creative writing | LMArena (Hugging Face dataset) | 139 | Claude Opus 5.51534 Elo |
| LMArena Text: English | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1538 Elo |
| LMArena Text: entertainment and media | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1526 Elo |
| LMArena Text: instruction following | LMArena (Hugging Face dataset) | 139 | Claude Opus 5.51536 Elo |
| LMArena Text: longer queries | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1546 Elo |
| LMArena Text: multi-turn | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1556 Elo |
| LMArena Text: non-English | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1522 Elo |
| LMArena Text: Russian | LMArena (Hugging Face dataset) | 139 | Gemini 4 Argon1534 Elo |
| LMArena Text: writing and language | LMArena (Hugging Face dataset) | 139 | Claude Opus 5.51534 Elo |
| LMArena Text: Chinese | LMArena (Hugging Face dataset) | 136 | Gemini 4 Argon1608 Elo |
| DTBench | Epoch AI benchmarking hub | 134 | Claude Opus 5.598.9% |
| IFBench | Artificial Analysis | 131 | Grok 4.383.3% |
| LMCA | Epoch AI benchmarking hub | 129 | Claude Sonnet 5.568.3% |
| LMArena Text: German | LMArena (Hugging Face dataset) | 123 | Claude Opus 51526 Elo |
| LMArena Text: Spanish | LMArena (Hugging Face dataset) | 123 | Claude Fable 5.11521 Elo |
| LMArena Text: French | LMArena (Hugging Face dataset) | 119 | Claude Fable 5.11527 Elo |
| LMArena Text: Korean | LMArena (Hugging Face dataset) | 115 | Claude Fable 5.11521 Elo |
| LMArena Text: Japanese | LMArena (Hugging Face dataset) | 113 | Claude Fable 5.11560 Elo |
| LMArena Text: Polish | LMArena (Hugging Face dataset) | 109 | Claude Fable 5.11524 Elo |
| LMArena Vision | LMArena (Hugging Face dataset) | 85 | Claude Fable 51325 Elo |
| LMArena Vision: English | LMArena (Hugging Face dataset) | 85 | Claude Opus 4.61324 Elo |
| MedScribe (Vals AI) | Vals AI boards | 85 | Claude Opus 5.591.4% |
| LMArena Vision: reading text | LMArena (Hugging Face dataset) | 76 | Claude Fable 51337 Elo |
| SAGE (Vals AI) | Vals AI boards | 74 | Claude Opus 4.756.1% |
| LMArena Vision: Chinese | LMArena (Hugging Face dataset) | 72 | Gemini 3.8 Flash1416 Elo |
| LMArena Vision: creative writing | LMArena (Hugging Face dataset) | 70 | Claude Fable 51338 Elo |
| LMArena Vision: humor | LMArena (Hugging Face dataset) | 62 | Claude Fable 51341 Elo |
| Public Benefits Bench (Vals AI) | Vals AI boards | 46 | Claude Opus 576.9% |
| LiveBench data analysis | LiveBench | 43 | Gemini 4 Argon85.2% |
| Furniture assembly | Epoch AI benchmarking hub | 33 | Claude Opus 5.583.3% |
| LMArena Vision: recognizing things | LMArena (Hugging Face dataset) | 33 | Gemini 3.1 Pro Preview1330 Elo |
| Vals Multimodal Index | Vals AI boards | 33 | Claude Fable 574.2% |
| LMArena Document | LMArena (Hugging Face dataset) | 29 | Claude Opus 51516 Elo |
| Fiction.LiveBench | Epoch AI benchmarking hub | 28 | GPT-597.2% |
| Surface Evolver Bench | Epoch AI benchmarking hub | 25 | Claude Fable 595.0% |
| LMArena Vision: captioning | LMArena (Hugging Face dataset) | 23 | Gemini 3.1 Pro Preview1345 Elo |
| Creative writing (Lech Mazur) | Epoch AI benchmarking hub | 22 | GPT-586.0% |
| Vals Index | Vals Index | 20 | Claude Sonnet 5.570.8% |
| CL-bench | Epoch AI benchmarking hub | 18 | GPT-5.427.9% |
| CL-bench Life | Epoch AI benchmarking hub | 13 | GPT-5.522.2% |
| ExploitBench | Epoch AI benchmarking hub | 8 | GPT-5.547.4% |
| WinoGrande | Epoch AI benchmarking hub | 4 | GPT-487.5% |
| HellaSwag | Epoch AI benchmarking hub | 3 | GPT-495.3% |
| PIQA | Epoch AI benchmarking hub | 3 | GPT-4o-mini88.7% |
Image generation
| Benchmark | Run by | Models measured | Leader today |
|---|---|---|---|
| Design Arena: graphic design | Design Arena | 7 | Nano Banana 2.11336 Elo |
| Design Arena: images | Design Arena | 7 | Nano Banana 2.11291 Elo |
| Design Arena: logos | Design Arena | 7 | Nano Banana 2.11278 Elo |
| LMArena Image Edit | LMArena (Hugging Face dataset) | 3 | Nano Banana 2.11428 Elo |
| LMArena Text to Image | LMArena (Hugging Face dataset) | 3 | Nano Banana 2.11328 Elo |