Epoch AI · GPQA diamond
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · FrontierMath-Tiers-1-3-v2-Private
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · FrontierMath-Tier-4-v2-Private
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · SimpleQA Verified
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · OTIS Mock AIME 2024-2025
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · MirrorCode
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · Chess Puzzles
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · Mystery Game Puzzles
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · Furniture Assembly
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
Epoch AI · Epoch Capabilities Index
Top 10 of the 30 we keep.
Source: Epoch AI (CC BY), as of Nov 20, 2026. Source
LMArena · Text (style control)
LMArena · Text (style control): top results| Rank | Model (as the source lists it) | Arena score |
|---|
| 1 | Gemini 4 Carbon (High) | 1,526 |
|---|
| 2 | Gemini 4 Argon (High) | 1,521 |
|---|
| 3 | Gemini 4 Neon (High) | 1,518 |
|---|
| 4 | Muse Wonder (Max) | 1,511 |
|---|
| 5 | Claude Opus 5.5 (High) | 1,507 |
|---|
| 6 | Claude Fable 5 (High) | 1,504 |
|---|
| 7 | Claude Opus 4.6 (High) | 1,503 |
|---|
| 8 | Claude Fable 5.1 (Max) | 1,502 |
|---|
| 9 | Claude Opus 4.7 (High) | 1,501 |
|---|
| 10 | Muse Spark 1.3 (Max) | 1,498 |
|---|
Top 10 of the 30 we keep.
LMArena (CC BY 4.0), as of Nov 18, 2026. Source
LMArena · WebDev
LMArena · WebDev: top results| Rank | Model (as the source lists it) | Arena score |
|---|
| 1 | Claude Fable 5.2 (xHigh) | 1,824 |
|---|
| 2 | Claude Opus 5.5 (Max) | 1,818 |
|---|
| 3 | GPT-6.2 Astra (Max) | 1,781 |
|---|
| 4 | Claude Sonnet 5.5 (xHigh) | 1,773 |
|---|
| 5 | Gemini 4 Carbon (High) | 1,769 |
|---|
| 6 | Gemini 4 Neon (High) | 1,764 |
|---|
| 7 | GPT-6 Astra (Max) | 1,763 |
|---|
| 8 | GPT-6.2 Sol (Max) | 1,757 |
|---|
| 9 | Qwen4-Max (xHigh) | 1,750 |
|---|
| 10 | GPT-6.1 Sol (Max) | 1,744 |
|---|
Top 10 of the 30 we keep.
LMArena (CC BY 4.0), as of Nov 18, 2026. Source
LMArena · Vision
LMArena · Vision: top results| Rank | Model (as the source lists it) | Arena score |
|---|
| 1 | Claude Fable 5 (High) | 1,326 |
|---|
| 2 | Claude Opus 5.5 (High) | 1,324 |
|---|
| 3 | Claude Fable 5.1 (Max) | 1,322 |
|---|
| 4 | Claude Opus 4.7 | 1,319 |
|---|
| 5 | Claude Opus 5 (High) | 1,317 |
|---|
| 6 | Gemini 3.8 Flash (High) | 1,317 |
|---|
| 7 | Claude Opus 4.6 (High) | 1,316 |
|---|
| 8 | Claude Opus 4.7 (High) | 1,314 |
|---|
| 9 | Gemini 3.7 Flash (High) | 1,313 |
|---|
| 10 | Claude Opus 4.6 | 1,312 |
|---|
Top 10 of the 30 we keep.
LMArena (CC BY 4.0), as of Nov 9, 2026. Source
LMArena · Agent
Top 10 of the 30 we keep.
LMArena (CC BY 4.0), as of Nov 18, 2026. Source
OpenRouter · GPQA Diamond
OpenRouter · GPQA Diamond: top results| Rank | Model (as the source lists it) | Accuracy |
|---|
| 1 | Gemini 4 Argonvia Google, its own provider | 96.5% |
|---|
| 2 | Gemini 3.8 Flashvia Google AI Studio, its own provider | 95.3% |
|---|
| 3 | Gemini 3.1 Pro Previewvia Google, its own provider | 95.2% |
|---|
| 4 | GPT-6 Astra Provia OpenAI, its own provider | 95.1% |
|---|
| 5 | Claude Fable 5.2via Anthropic, its own provider | 95.1% |
|---|
| 6 | Ling 3.1 Flashvia DeepInfra, the most accurate provider | 94.8% |
|---|
| 7 | GPT-6 Astravia OpenAI, its own provider | 94.7% |
|---|
| 8 | Gemini 3.7 Flashvia Google, its own provider | 94.5% |
|---|
| 9 | GPT-6.1 Solvia OpenAI, its own provider | 94.4% |
|---|
| 10 | GPT-5.6 Sol Provia OpenAI, its own provider | 93.9% |
|---|
Top 10 of the 60 we keep.
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Nov 20, 2026. Source
OpenRouter · GPQA Diamond: cost per task
Lower is better on this board.
OpenRouter · GPQA Diamond: cost per task: top results| Rank | Model (as the source lists it) | Cost per task |
|---|
| 1 | GPT-6 Lunavia OpenAI, its own provider | $0.003 |
|---|
| 2 | Claude Haiku 5.5via Anthropic, its own provider | $0.004 |
|---|
| 3 | MiniMax M3.1 Flashvia Minimax, its own provider | $0.0068 |
|---|
| 4 | DeepSeek V4 Flash 0423via Crusoe, the most accurate provider | $0.008 |
|---|
| 5 | GPT-6 Luna Provia OpenAI, its own provider | $0.0082 |
|---|
| 6 | Ling 3.1 Flashvia DeepInfra, the most accurate provider | $0.0093 |
|---|
| 7 | GLM 5.3 Flashvia Z.AI, its own provider | $0.01 |
|---|
| 8 | GPT-6.1 Terravia OpenAI, its own provider | $0.011 |
|---|
| 9 | DeepSeek V4.1 Flash 1017via Tenstorrent, the most accurate provider | $0.012 |
|---|
| 10 | Nemotron 3.5 Super VLvia Io Net, the most accurate provider | $0.013 |
|---|
Top 10 of the 60 we keep.
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Nov 20, 2026. Source
OpenRouter · GPQA Diamond: time per task
Lower is better on this board.
OpenRouter · GPQA Diamond: time per task: top results| Rank | Model (as the source lists it) | Time per task |
|---|
| 1 | Claude Sonnet 5.5via Anthropic, its own provider | 20 s |
|---|
| 2 | Claude Fable 5.2via Anthropic, its own provider | 26 s |
|---|
| 3 | Claude Opus 5.5via Anthropic, its own provider | 27 s |
|---|
| 4 | GPT-6.1 Terravia OpenAI, its own provider | 28 s |
|---|
| 5 | GPT-6 Solvia OpenAI, its own provider | 34 s |
|---|
| 6 | Claude Haiku 5.5via Anthropic, its own provider | 37 s |
|---|
| 7 | Claude Fable 5.1via Anthropic, its own provider | 40 s |
|---|
| 8 | Claude Fable 5via Anthropic, its own provider | 47 s |
|---|
| 9 | GPT-6.1 Solvia OpenAI, its own provider | 48 s |
|---|
| 10 | GPT-6 Astravia OpenAI, its own provider | 48 s |
|---|
Top 10 of the 60 we keep.
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Nov 20, 2026. Source
OpenRouter · τ²-bench airline
OpenRouter · τ²-bench airline: top results| Rank | Model (as the source lists it) | Accuracy |
|---|
| 1 | DeepSeek V4 Flash 0423via OpenInference, the most accurate provider | 81.7% |
|---|
| 2 | DeepSeek V4 Flash 0731via BaseTen, the most accurate provider | 79.8% |
|---|
| 3 | Qwen3.5 397B A17Bvia Alibaba, its own provider | 79.4% |
|---|
| 4 | Ling 3.1 Flashvia DeepInfra, the most accurate provider | 79.3% |
|---|
| 5 | Claude Fable 5via Anthropic, its own provider | 79.3% |
|---|
| 6 | Qwen3.5-35B-A3Bvia Alibaba, its own provider | 79.2% |
|---|
| 7 | Claude Fable 5.1via Anthropic, its own provider | 79.1% |
|---|
| 8 | MiMo-V2.6-Provia Xiaomi, its own provider | 78.9% |
|---|
| 9 | Gemini 4 Argonvia Google, its own provider | 78.7% |
|---|
| 10 | DeepSeek V4 Pro 0813via DeepInfra, the most accurate provider | 78.6% |
|---|
Top 10 of the 59 we keep.
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Nov 20, 2026. Source
OpenRouter · τ²-bench airline: cost per task
Lower is better on this board.
OpenRouter · τ²-bench airline: cost per task: top results| Rank | Model (as the source lists it) | Cost per task |
|---|
| 1 | Ling 3.0 Flashvia Novita, the most accurate provider | $0.0021 |
|---|
| 2 | DeepSeek V4 Flash 0731via BaseTen, the most accurate provider | $0.0076 |
|---|
| 3 | MiMo-V2.5via Xiaomi, its own provider | $0.0082 |
|---|
| 4 | GLM 5.3 Flashvia Z.AI, its own provider | $0.0092 |
|---|
| 5 | MiMo-V2.6-Flashvia Xiaomi, its own provider | $0.012 |
|---|
| 6 | Qwen3.8 Flashvia Alibaba, its own provider | $0.015 |
|---|
| 7 | GLM 5.3via digitalocean, the most accurate provider | $0.016 |
|---|
| 8 | DeepSeek V4.1 Flash 1017via AtlasCloud, the most accurate provider | $0.017 |
|---|
| 9 | Beam 1.0 501B A23Bvia Io Net, the most accurate provider | $0.025 |
|---|
| 10 | MiniMax M3via Minimax, its own provider | $0.03 |
|---|
Top 10 of the 60 we keep.
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Nov 20, 2026. Source
OpenRouter · τ²-bench airline: time per task
Lower is better on this board.
OpenRouter · τ²-bench airline: time per task: top results| Rank | Model (as the source lists it) | Time per task |
|---|
| 1 | DeepSeek V4 Flash 0731via BaseTen, the most accurate provider | 31 s |
|---|
| 2 | GLM 5.3via digitalocean, the most accurate provider | 42 s |
|---|
| 3 | Muse Spark Openvia Modal, the most accurate provider | 66 s |
|---|
| 4 | DeepSeek V4 Pro 0423via BaseTen, the most accurate provider | 70 s |
|---|
| 5 | Claude Sonnet 5.5via Anthropic, its own provider | 73 s |
|---|
| 6 | Ling 3.0 Flashvia Novita, the most accurate provider | 76 s |
|---|
| 7 | Claude Opus 4.7via Anthropic, its own provider | 86 s |
|---|
| 8 | DeepSeek V4.1 Flash 1017via AtlasCloud, the most accurate provider | 93 s |
|---|
| 9 | Claude Sonnet 5via Anthropic, its own provider | 95 s |
|---|
| 10 | Nova Micro 1.0via Anthropic, the most accurate provider | 96 s |
|---|
Top 10 of the 60 we keep.
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Nov 20, 2026. Source
Vectara · Hallucination rate in summaries
Lower is better on this board.
Top 10 of the 73 we keep.
Vectara Hallucination Leaderboard (HHEM-2.3), Apache-2.0, as of Oct 22, 2026. Source