Arena Vision Style-Controlled
What it measures and how
Read only overall from the official vision_style_control dataset. Excludes image generation and editing.
How this evidence is used
Receives the 6% vision budget. This is still human-preference evidence, not strict image accuracy.
Benchmark results
Fixed rules select one representative configuration per public model. Anonymous test models are omitted. Original source ranks are retained, with up to 30 models per benchmark.
| Source rank | Source model name | Raw result | Representative configuration |
|---|---|---|---|
| 1 | claude-fable-5-highAnthropic | 1,308 | High reasoning |
| 2 | qwen3.8-maxAlibaba | 1,300.7 | Source default configuration |
| 3 | claude-opus-4-6-highAnthropic | 1,299.2 | High reasoning |
| 5 | claude-opus-4-7-highAnthropic | 1,297.8 | High reasoning |
| 6 | gemini-3.7-flash-highGoogle | 1,296.9 | High reasoning |
| 8 | gpt-6.1-sol-maxOpenAI | 1,294.1 | Max reasoning |
| 9 | muse-sparkMeta | 1,293.9 | Source default configuration |
| 10 | muse-spark-1.2 (xHigh)Meta | 1,292.9 | xHigh reasoning |
| 11 | claude-opus-5.5-highAnthropic | 1,292.7 | High reasoning |
| 12 | muse-spark-1.3-maxMeta | 1,289.7 | Max reasoning |
| 13 | gemini-3-proGoogle | 1,289.3 | Source default configuration |
| 14 | gemini-3.8-flash-highGoogle | 1,289.1 | High reasoning |
| 15 | claude-opus-5-highAnthropic | 1,287.4 | High reasoning |
| 16 | gpt-5.6-sol-xhighOpenAI | 1,287 | xHigh reasoning |
| 17 | claude-fable-5.1-maxAnthropic | 1,286.4 | Max reasoning |
| 19 | gpt-6-astra-maxOpenAI | 1,286 | Max reasoning |
| 21 | claude-opus-4-8-highAnthropic | 1,285.6 | High reasoning |
| 23 | gpt-5.4-highOpenAI | 1,285 | High reasoning |
| 24 | gemini-3.5-flash-highGoogle | 1,283.7 | High reasoning |
| 25 | gpt-5.5-highOpenAI | 1,281.1 | High reasoning |
| 26 | gemini-3.6-flash-highGoogle | 1,280.7 | High reasoning |
| 27 | grok-4.5xAI | 1,280 | Source default configuration |
| 28 | muse-spark-1.1Meta | 1,280 | Source default configuration |
| 29 | gemini-3.1-pro-previewGoogle | 1,279.8 | Source default configuration |
| 32 | gpt-5.2-chat-latest-20260210OpenAI | 1,277.9 | Source default configuration |
| 33 | gpt-5.5-instantOpenAI | 1,275.7 | Source default configuration |
| 34 | claude-sonnet-4-6Anthropic | 1,275 | Source default configuration |
| 35 | glm-5.3-flashZ.ai | 1,273.3 | Source default configuration |
| 36 | gemini-3-flashGoogle | 1,272.8 | Source default configuration |
| 37 | gpt-6-sol-maxOpenAI | 1,270.4 | Max reasoning |
Limitations and attribution
Human preference depends on style, user demographics and prompt distribution. Overall and creative categories have overlapping votes and share the original human-preference budget. They do not count as two independent organizations.
Data license: CC BY 4.0
Results are published by LMArena. Raw scores and the e/acc consensus index use different scales and cannot be added directly.