Arena WebDev
LMArena · Human blind preference tests whether websites are usable and ready to deliver.
Across e/accCounts toward ranking
Evidence budget2.4%
Upstream data time10/13 08:00
Last successful sync10/17 08:05
What it measures and how
Use only the WebDev overall board in official published snapshots. Do not scrape the live arena.ai page. Together with Text, it forms one Arena family vote.
How this evidence is used
Receives 2.4% from the coding and design budget. In the aesthetics category it measures preference for websites, not code accuracy. It still shares the evidence family with other Arena tasks.
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-opus-5.5-maxAnthropic | 1,811.1 | Max reasoning |
| 2 | gpt-6-astra-maxOpenAI | 1,780.2 | Max reasoning |
| 3 | claude-sonnet-5.5-xhighAnthropic | 1,773.7 | xHigh reasoning |
| 4 | gpt-6.1-sol-maxOpenAI | 1,760.1 | Max reasoning |
| 5 | claude-fable-5.1-maxAnthropic | 1,742.7 | Max reasoning |
| 7 | claude-opus-5-maxAnthropic | 1,690 | Max reasoning |
| 8 | gpt-6-sol-maxOpenAI | 1,685.9 | Max reasoning |
| 9 | gemini-4-argon-highGoogle | 1,678 | High reasoning |
| 10 | qwen3.8-max-0902Alibaba | 1,673.2 | Source default configuration |
| 11 | qwen3.8-maxAlibaba | 1,672 | Source default configuration |
| 12 | muse-spark-1.3-maxMeta | 1,658.5 | Max reasoning |
| 14 | kimi-k3-maxMoonshot AI | 1,655.3 | Source default configuration |
| 15 | muse-wonder-maxMeta | 1,650.9 | Max reasoning |
| 16 | grok-4.7-xhighxAI | 1,640.1 | xHigh reasoning |
| 17 | mimo-v2.6-flashXiaomi | 1,635.1 | Source default configuration |
| 18 | qwen3.8-flash-nextAlibaba | 1,633.2 | Source default configuration |
| 19 | hy4-previewTencent | 1,631.6 | Source default configuration |
| 20 | mimo-v2.6-proXiaomi | 1,625.2 | Source default configuration |
| 21 | claude-fable-5-highAnthropic | 1,623.8 | High reasoning |
| 23 | glm-5.3-maxZ.ai | 1,621.2 | Source default configuration |
| 24 | deepseek-v4.1-flash-maxDeepSeek | 1,618.5 | Source default configuration |
| 25 | grok-4.6-highxAI | 1,617.4 | High reasoning |
| 26 | gpt-5.6-sol-xhigh (codex-harness)OpenAI | 1,617 | xHigh reasoning · codex-harness |
| 27 | glm-5.3-flashZ.ai | 1,606.1 | Source default configuration |
| 28 | glm-5.2-maxZ.ai | 1,604.1 | Source default configuration |
| 29 | qwen3.8-27bAlibaba | 1,594.2 | Source default configuration |
| 30 | gemini-3.7-flash-highGoogle | 1,589.9 | High reasoning |
| 31 | claude-haiku-5.5-highAnthropic | 1,585 | High reasoning |
| 32 | gemini-3.8-flash-highGoogle | 1,584.8 | High reasoning |
| 33 | deepseek-v4-pro-high-20260813DeepSeek | 1,582.2 | Source default configuration |
Limitations and attribution
Its preference construct is correlated with Text.
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.