LMArena Image-to-WebDev
2026-10-09 · Frontend
- What it measures
- Generate a web page from an image or screenshot. The source page also includes agentic coding flows that need several steps and tool use.
- How
- Two anonymous models each build a page from the same image. A person votes after looking. Only votes cast before the identities are revealed count.
- What the score means
- The score is a pairwise strength estimate, read like Elo. The +30/-30 beside a score is the interval printed on the page, from that row’s rating bounds.
- Caveats
- It measures whether a person prefers the generated page, not whether tests pass. Effort tiers and harnesses of the same model are ranked separately. Preliminary rows have fewer votes and can move. This is not the text-prompt WebDev board.
- Updates
- Updated continuously. This page's board date is 2026-10-09, synced 2026-10-11. We sync weekly and write changes on the home page.
Full board
64
All
| # | Rank spread | Model | Score | CI | Votes | Price $/M | Context |
|---|
| 1 | 1–5 | claude-opus-5.5-maxAnthropic · Proprietarymax | 1749 | +30/-30 | 493 | $4 / $20 | 1M |
| 2 | 1–5 | claude-sonnet-5.5-xhighAnthropic · Proprietaryxhigh | 1740 | +20/-20 | 1,074 | $2 / $10 | 1M |
| 3 | 1–5 | gpt-6-astra-maxOpenAI · Proprietarymax | 1731 | +13/-13 | 2,739 | $10 / $50 | 1.1M |
| 4 | 1–5 | claude-fable-5.1-maxAnthropic · Proprietarymax | 1720 | +14/-14 | 2,470 | $10 / $50 | 1M |
| 5 | 1–5 | gpt-6.1-sol-maxOpenAI · Proprietarymax | 1703 | +17/-17 | 1,430 | $2 / $10 | 1.1M |
| 6 | 6–9 | gpt-6-sol-maxOpenAI · Proprietarymax | 1661 | +13/-13 | 2,426 | $2 / $10 | 1.1M |
| 7 | 6–9 | claude-opus-5-maxAnthropic · Proprietarymax | 1661 | +10/-10 | 4,250 | $5 / $25 | 1M |
| 8 | 6–9 | muse-spark-1.3-maxMeta · Proprietarymax | 1642 | +10/-10 | 4,688 | $1.25 / $4.25 | 1M |
| 9 | 6–12 | qwen3.8-max-0902Alibaba · Proprietary0902Preliminary | 1638 | +18/-18 | 1,164 | $2 / $6 | 1M |
| 10 | 9–14 | claude-fable-5-highAnthropic · Proprietaryhigh | 1622 | +9/-9 | 5,457 | $10 / $50 | 1M |
| 11 | 9–19 | qwen3.8-maxAlibaba · ProprietaryPreliminary | 1614 | +10/-10 | 3,971 | $2 / $6 | 1M |
| 12 | 9–23 | deepseek-v4.1-flash-maxDeepSeek · MITmax | 1609 | +22/-22 | 766 | $0.30 / $1.20 | 1M |
| 13 | 10–20 | gpt-5.6-sol-xhigh (codex-harness)OpenAI · Proprietaryxhigh | 1608 | +10/-10 | 4,478 | $4 / $20 | 1.1M |
| 14 | 10–21 | gemini-3.7-flash-highGoogle · ProprietaryPreliminary | 1603 | +13/-13 | 2,144 | $0.75 / $3.75 | 1M |
| 15 | 11–22 | grok-4.6-highSpaceXAI · Proprietary | 1601 | +12/-12 | 2,826 | $2 / $6 | 500K |
| 16 | 11–24 | muse-spark-1.3 (xHigh)Meta · Proprietaryxhigh | 1598 | +14/-14 | 2,152 | $1.25 / $4.25 | 1M |
| 17 | 11–24 | mimo-v2.6-flashXiaomi · MIT | 1597 | +14/-14 | 1,930 | $0.14 / $0.28 | 1.1M |
| 18 | 11–25 | mimo-v2.6-proXiaomi · MIT | 1595 | +13/-13 | 2,199 | $0.43 / $0.87 | 1.1M |
| 19 | 11–25 | gemini-3.8-flash-highGoogle · ProprietaryhighPreliminary | 1593 | +14/-14 | 2,015 | $0.75 / $3.75 | 1M |
| 20 | 13–26 | glm-5.3-flashZ.ai · MIT | 1581 | +10/-10 | 4,315 | $0.06 / $0.20 | 1M |
Score and price
The horizontal axis is the Pareto page’s blended price, (input + 3×output) / 4, in $/M. The vertical axis is the board score. Filled points are the six models that page lists, with the prices printed there.
arena.ai/leaderboard/code/image-to-webdev/pareto- claude-opus-5.5-max1749 · $16/M
- claude-sonnet-5.5-xhigh1740 · $8/M
- muse-spark-1.3-max1642 · $3.50/M
- deepseek-v4.1-flash-max1609 · $0.97/M
- mimo-v2.6-flash1597 · $0.25/M
- glm-5.3-flash1581 · $0.17/M