Arena vs Artificial Analysis
The two leaderboards people quote most rank the AI labs differently. Here is each lab's best model on both, the rank gap, and how strongly the two lists agree.
Data as of the week of 26 Oct 2026 (updated 26 Oct 2026, 08:09, Bucharest time). Every figure links to its source; see the methodology.
Lab by lab: Arena rank vs AA position
| Lab | Arena (rank among these labs) | Artificial Analysis (rank among these labs) | Who rates it higher | Same model on both? |
|---|---|---|---|---|
| Anthropic | #2 (#2 of all, 1509) Claude Opus 5.5 (High) | #1 (#1 of all, 57.6) Claude Opus 5.5 (Max, Default Fallback) | AA +1 | same model |
| Google DeepMind | #1 (#1 of all, 1517) Gemini 4 Argon (High) | #5 (#10 of all, 52.6) Gemini 4 Argon (High) | Arena +4 | same model |
| OpenAI | #5 (#21 of all, 1485) GPT-5.6 Sol (xHigh) | #4 (#9 of all, 52.7) GPT-6 Astra (Max) | AA +1 | different models |
| Meta Superintelligence Labs | #3 (#3 of all, 1509) Muse Wonder (Max) | #2 (#4 of all, 55.8) Muse Wonder Full (Max) | AA +1 | same model |
| xAI (SpaceXAI) | #9 (#39 of all, 1475) Grok 4.20 Beta 1 | #6 (#25 of all, 46.4) Grok 4.7 (Xhigh) | AA +3 | different models |
| Alibaba (Qwen) | #6 (#23 of all, 1483) Qwen3.8-Max | #7 (#29 of all, 45.4) Qwen3.8 Max (0902) | Arena +1 | same model |
| Moonshot AI (Kimi) | #4 (#18 of all, 1487) Kimi K3 (Max) | #3 (#5 of all, 55.1) Kimi K3.1 Pro | AA +1 | same model |
| Z.ai (Zhipu) | #7 (#26 of all, 1481) GLM-5.3 (Max) | #9 (#31 of all, 44.8) GLM-5.3 (Max) | Arena +2 | same model |
| DeepSeek | #8 (#34 of all, 1476) DeepSeek-V4.1-Flash-20260910 (Max) | #8 (#30 of all, 45.1) DeepSeek V4.1 Flash 1017 (Max) | agree | different models |
| Mistral AI | #10 (#111 of all, 1434) Mistral Large 4 | #10 (#49 of all, 38.4) Mistral Large 4 Preview | agree | same model |
Ranks among the 10 labs (computed by us) from each lab's best model: Arena text leaderboard (human preference votes, manual snapshot 22 Oct 2026) and the Artificial Analysis Intelligence Index (benchmarks, fetched 26 Oct 2026, 08:08). "Same model" is a name match ignoring punctuation, the effort setting in brackets and trailing version suffixes.
How much do they agree?
Why they differ: Arena ranks models by blind human votes on chat answers; Artificial Analysis averages scores on a fixed set of benchmarks. Read the longer explanation in why Arena and AA disagree.
FAQ
Do Arena and Artificial Analysis agree on the best AI lab?
No: Google DeepMind leads Arena, Anthropic leads Artificial Analysis.
Where do they disagree most?
Google DeepMind (Arena ranks it 4 places higher); xAI (SpaceXAI) (AA ranks it 3 places higher); Z.ai (Zhipu) (Arena ranks it 2 places higher).
Which one does the scoreboard use?
Both. Each is one input to the composite score with its own weight; see the methodology.