Best AI for building websites: 148 models ranked by blind votes
Two models build a working web page or app from the same prompt, and voters pick the better result. Ranked by LMArena’s blind votes, every score with its margin of error and, beside it, the list price of each model on our price list.
Scores from LMArena, published 19 October 2026, under CC BY 4.0. Prices from each maker’s pricing page, read on 15 October 2026.
- Models ranked
- 148
- Level with the leader
- 1
- On our price list
- 0
- Published
- 19 October 2026
Level: the votes cannot separate it from the leader.
On LMArena’s Web development board, Claude Opus 5.5 (Max) from Anthropic leads with 1807 points from 2,925 blind votes, published 19 October 2026. No other model is statistically level with it. None of the models it ranks has an API price on our list yet.
Best AI for building websites
Two models build a working web page or app from the same prompt, and voters pick the better result.
| # | Model | Score | Votes | Per 1M tokens, in / out |
|---|---|---|---|---|
| 1 | Claude Opus 5.5 (Max)Anthropic | 95% interval 1794 to 1821 | 2,925 | · |
| 2 | GPT-6 Astra (Max)OpenAI | 95% interval 1764 to 1784 | 6,169 | · |
| 2 | Claude Sonnet 5.5 (xHigh)Anthropic | 95% interval 1756 to 1781 | 3,428 | · |
| 2 | GPT-6.1 Sol (Max)OpenAI | 95% interval 1740 to 1766 | 3,072 | · |
| 4 | Claude Fable 5.1 (Max)Anthropic | 95% interval 1734 to 1752 | 7,744 | · |
| 4 | Muse Wonder Full (Max)Meta | 95% interval 1710 to 1742 | 1,639 | · |
| 6 | Claude Sonnet 5.5 (High)Anthropic | 95% interval 1700 to 1720 | 5,158 | · |
| 8 | Claude Opus 5 (Max)Anthropic | 95% interval 1683 to 1695 | 18,555 | · |
| 8 | GPT-6 Sol (Max)OpenAI | 95% interval 1673 to 1693 | 4,608 | · |
| 8 | Gemini 4 Argon (High)Google | 95% interval 1663 to 1694 | 1,728 | · |
| 8 | Qwen3.8-MaxAlibaba | 95% interval 1660 to 1684 | 3,428 | · |
| 9 | Qwen3.8-Max (0902)Alibaba | 95% interval 1667 to 1681 | 14,421 | · |
| 9 | DeepSeek-V4.1-Flash-20261017 (Max)DeepSeek · open weights | 95% interval 1650 to 1679 | 2,008 | · |
| 11 | Muse Spark 1.3 (Max)Meta | 95% interval 1651 to 1667 | 9,407 | · |
| 11 | Claude Opus 5 (High)Anthropic | 95% interval 1653 to 1664 | 24,739 | · |
| 11 | Muse Wonder (Max)Meta | 95% interval 1638 to 1665 | 2,289 | · |
| 12 | Kimi K3 (Max)Moonshot AI · open weights | 95% interval 1649 to 1661 | 19,592 | · |
| 17 | Grok 4.7 (xHigh)xAI | 95% interval 1627 to 1647 | 5,303 | · |
| 17 | MiMo-V2.6-FlashXiaomi · open weights | 95% interval 1624 to 1643 | 5,293 | · |
| 17 | Hy4 previewTencent · open weights | 95% interval 1621 to 1641 | 4,798 | · |
| 18 | Qwen3.8-Flash-NextAlibaba · open weights | 95% interval 1621 to 1637 | 8,382 | · |
| 18 | MiMo-V2.6-ProXiaomi · open weights | 95% interval 1615 to 1635 | 5,041 | · |
| 18 | Claude Fable 5 (High)Anthropic | 95% interval 1618 to 1630 | 15,899 | · |
| 18 | Muse Spark 1.3 (xHigh)Meta | 95% interval 1614 to 1629 | 8,767 | · |
| 18 | GLM-5.3 (Max)Z.ai · open weights | 95% interval 1614 to 1629 | 9,828 | · |
| 18 | DeepSeek-V4.1-Flash-20260910 (Max)DeepSeek · open weights | 95% interval 1607 to 1629 | 3,528 | · |
| 20 | Grok 4.6 (High)xAI | 95% interval 1609 to 1624 | 10,483 | · |
| 20 | GPT-5.6 Sol (xHigh, Codex Harness)OpenAI | 95% interval 1610 to 1622 | 17,786 | · |
| 26 | GLM-5.3-FlashZ.ai · open weights | 95% interval 1600 to 1613 | 18,264 | · |
| 26 | GLM-5.2 (Max)Z.ai · open weights | 95% interval 1598 to 1610 | 16,657 | · |
| 29 | MiniMax-M3.1-FlashMiniMax · open weights | 95% interval 1573 to 1602 | 1,853 | · |
| 30 | Qwen3.8-27BAlibaba · open weights | 95% interval 1586 to 1599 | 16,385 | · |
| 31 | Gemini 3.7 Flash (High)Google | 95% interval 1584 to 1598 | 12,636 | · |
| 31 | Claude Haiku 5.5 (High)Anthropic | 95% interval 1575 to 1597 | 3,514 | · |
| 31 | Gemini 3.8 Flash (High)Google | 95% interval 1577 to 1592 | 10,851 | · |
| 31 | DeepSeek-V4-Pro (High, 20260813)DeepSeek · open weights | 95% interval 1573 to 1592 | 4,714 | · |
| 31 | DeepSeek-V4-Flash (High)DeepSeek · open weights | 95% interval 1573 to 1591 | 4,947 | · |
| 33 | GPT-6 Luna (Max)OpenAI | 95% interval 1572 to 1584 | 21,622 | · |
| 34 | Step 5 Preview (High)StepFun | 95% interval 1557 to 1577 | 4,728 | · |
| 39 | Claude Opus 4.7 (High)Anthropic | 95% interval 1552 to 1564 | 17,910 | · |
| 39 | Claude Opus 4.8 (High)Anthropic | 95% interval 1550 to 1562 | 18,493 | · |
| 39 | Claude Opus 4.7Anthropic | 95% interval 1550 to 1561 | 18,230 | · |
| 39 | Grok 4.5xAI | 95% interval 1546 to 1560 | 11,082 | · |
| 40 | Mistral Large 4Mistral AI | 95% interval 1530 to 1555 | 2,832 | · |
| 41 | Claude Opus 4.6 (High)Anthropic | 95% interval 1541 to 1552 | 19,383 | · |
| 43 | Muse Spark 1.1Meta | 95% interval 1535 to 1550 | 7,985 | · |
| 43 | Claude Sonnet 5 (High)Anthropic | 95% interval 1535 to 1546 | 15,755 | · |
| 44 | Gemini 3.6 Flash (High)Google | 95% interval 1531 to 1546 | 8,762 | · |
| 44 | Claude Opus 4.6Anthropic | 95% interval 1532 to 1543 | 20,644 | · |
| 44 | Claude Opus 4.8Anthropic | 95% interval 1530 to 1542 | 17,398 | · |
Where the ranking comes from
The scores, intervals and ranks are LMArena’s. The prices are ours.
- Blind votes
- LMArena puts the same prompt to two anonymous models and asks the person who wrote it to pick the better answer. The votes place every model on one scale, where a gap of 100 points means the higher model wins about 64 times in 100.
- Shared ranks
- A model’s rank is one plus the number of models whose whole 95% interval sits above its own. Models the votes cannot separate share a rank, which is why several can be first.
- Settings
- A model can appear more than once, at different reasoning settings: the line under its name gives the entry as LMArena lists it. Entries with web search switched on are a different product and carry no price.
- Prices
- The list price per million tokens from the maker’s own pricing page, read on 15 October 2026. It is a price per token, not per answer: a higher reasoning setting spends more tokens on the same question.
Source: LMArena, leaderboard dataset (revision 9041150e, published 19 October 2026), licensed under CC BY 4.0. We change no score: we add the prices and the links to our model pages.
What each of them costs to run
The price list works three everyday jobs on every model, and our own benchmark runs them on real business tasks to see which pass.
Scores published 19 October 2026