Project AURIONAI Use-Cases Research & Innovation
LMArena (Arena)

Arena Vision leaderboard

Crowd-sourced pairwise preference on vision chat

human preference arenaactive

Score in Elo, Arena Vision overall, style control default, 165 models

  1. 1AnthropicClaude Fable 51,308 Eloindependent
  2. 2Alibaba (Qwen)Qwen3.8 Max1,301 Eloindependent
  3. 3AnthropicClaude Opus 4.61,299 Eloindependent
14scores
13models
14independent
0maintainer-run
0vendor-reported

Definition

Version
not in the notes
Category
human preference arena
Maintainer
LMArena (Arena)
Status
active
Home page
https://arena.ai/leaderboard/chat/vision

Sources

  1. Vision Arena: VLM Leaderboard & Best Vision Models arena.ai https://arena.ai/leaderboard/chat/vision
  2. Arena Leaderboard: Official AI Model Rankings & Benchmarks arena.ai https://arena.ai/leaderboard

Scores at a Glance

Score in Elo, Arena Vision overall, style control default, 165 models

Higher is better for this unit. Elo has no zero, so the axis starts at 1,270.

1AnthropicClaude Fable 5Anthropichigh
1,308 Elo
independent2Alibaba (Qwen)Qwen3.8 MaxAlibaba (Qwen)
1,301 Elo
independent3AnthropicClaude Opus 4.6Anthropichigh
1,299 Elo
independent4Google DeepMindGemini 3.7 FlashGoogle DeepMindhigh; preliminary
1,297 Elo
independent5MetaMuse SparkMetamuse-spark
1,294 Elo
independent6OpenAIGPT-6.1 SolOpenAImax
1,294 Elo
independent7AnthropicClaude Opus 5.5Anthropichigh
1,293 Elo
independent8MetaMuse Spark 1.2MetaxHigh
1,293 Elo
independent9MetaMuse Spark 1.3Metamax (muse-spark-1.3-max)
1,290 Elo
independent10Google DeepMindGemini 3.8 FlashGoogle DeepMindhigh; preliminary
1,289 Elo
independent11AnthropicClaude Fable 5.1Anthropicmax
1,286 Elo
independent12OpenAIGPT-6 AstraOpenAImax
1,286 Elo
independent13MetaMuse Spark 1.1Meta
1,280 Elo
independent
1,2701,2801,2901,3001,310
Where this comes from
Each bar is one model, with its most independent recorded score. Hatched bars are vendor-reported; the word beside each bar says who ran it. Settings and dates differ between rows; the Scores table below has every result.

Scores

14 of 14 scores shown lmarena-vision-scores.csv
ModelScoreWho ran itSettingView or populationDateSource
AnthropicClaude Fable 51,308 Elo ±7independenthighrank 1 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
Alibaba (Qwen)Qwen3.8 Max1,301 Elo ±7independent–rank 2 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
AnthropicClaude Opus 4.61,299 Elo ±6independenthighrank 3 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
Google DeepMindGemini 3.7 Flash1,297 Elo ±10independenthigh; preliminaryrank 6 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
MetaMuse Spark1,294 Elo ±9independentmuse-sparkrank 9 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
OpenAIGPT-6.1 Sol1,294 Elo ±14independentmaxrank 8 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
AnthropicClaude Opus 5.51,293 Elo ±13independenthighrank 11 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
MetaMuse Spark 1.21,293 Elo ±14independentxHighrank 10 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
MetaMuse Spark 1.31,290 Elo ±10independentmax (muse-spark-1.3-max)rank 12 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
Google DeepMindGemini 3.8 Flash1,289 Elo ±11independenthigh; preliminaryrank 14 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
AnthropicClaude Fable 5.11,286 Elo ±10independentmaxrank 17 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
MetaMuse Spark 1.31,286 Elo ±10independentxHighrank 18 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
OpenAIGPT-6 Astra1,286 Elo ±11independentmaxrank 19 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
MetaMuse Spark 1.11,280 Elo ±7independent–rank 28 of 165, Arena Vision overall, style control default, 165 models9 Oct 2026arena.ai
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Arena Vision Leaderboard", https://aurion.semai.id/benchmarks/lmarena-vision/, records of October 2026.