Anthropic's Claude Opus 5.5 retains top spot

In November 2026, Anthropic's Claude Opus 5.5 remained the leading model in the WebDev Arena, a benchmark that evaluates artificial intelligence models based on anonymized duels. This ranking is updated regularly and reflects the performance of models in tasks related to web and software development.

Competitors are catching up

Despite Anthropic's continued dominance, other major players like OpenAI, Google, Meta, and Moonshot are making progress. OpenAI's GPT-6 Astra, one of the most advanced models from the company, was recently outperformed by Claude Opus 5.5. Google's Gemini 4 Carbon, currently in testing with a limited group of cyberdefense experts, finished third in the latest rankings.

Some models drop out of the top 10

Gemini 4 Argon, which was in ninth place last month, is no longer among the top 10 models. This version of Google's Gemini was recently made available to paying subscribers but has since fallen behind. Meta's Muse Wonder Full, which became available in October through its API and Muse Code, reached seventh place, just below Claude Fable 5.1.

Moonshot AI enters the top 10

Moonshot AI's Kimi K3.1 Pro also made it into the top 10, securing ninth place. Anthropic continues to be the best-represented company in the rankings, with four of its models occupying positions in the top 10. The complete list includes models such as Claude Opus 5.5, GPT-6 Astra, and Gemini 4 Carbon, each with their respective Elo scores.

OpenAI leads in fullstack development

In the category of fullstack development, OpenAI's GPT-6 Astra and GPT-6.1 Sol models took the top two spots. These models outperformed Anthropic's Claude Opus 5.5, which was ranked third in this specific area. However, these rankings are based on fewer votes compared to the overall WebDev Arena rankings, and the results should be interpreted with caution.

Anthropic dominates in front-end tasks

In the front-end development category, which assesses models on multi-step reasoning, tool calling, and HTML or React code generation, Anthropic's models lead. Claude Opus 5.5 is the top performer, followed by Claude Sonnet 5.5 and Google's Gemini 4 Carbon. These rankings, however, are based on a smaller number of votes than the general ranking.

WebDev Arena's ranking system

The WebDev Arena evaluates AI models using anonymized duels, where each model receives the same instruction, and users vote on which result seems most complete. The models are assigned Elo scores, which change over time based on the outcomes of these duels. A model can gain points by defeating a higher-ranked opponent or lose points by losing to a lower-ranked one. The top-ranked model can still gain points, but at a slower rate.