Anthropic Outperforms WebDevArena, Google and OpenAI Left Behind

⚡

Key Takeaways

1Anthropic dominates the WebDevArena in May 2026 with six models in the top 10, confirming its supremacy.

2Anthropic's Claude Opus 4.6 and 4.7 occupy four of the top five spots in the ranking.

3Google and OpenAI are falling behind, with OpenAI's GPT-5.5 dropping to the tenth position, behind Meta and its Muse Spark.

💡Why it matters — Anthropic's dominance highlights a shift in leadership in AI models for web development, impacting the competition.

Anthropic Continues to Dominate the WebDevArena

As of May 2026, Anthropic maintains its dominant position in the WebDevArena, a ranking by LMArena that evaluates the performance of AI models on front-end web development tasks. These tasks include agentic coding workflows that require multi-step reasoning and the use of tools. Anthropic has placed six of its models in the top 10, notably occupying four of the top five positions thanks to the releases of Claude Opus 4.6 and 4.7.

Claude Opus 4.7 Tops the Rankings

The standard and "thinking" versions of Claude Opus 4.7, deployed on April 16, logically take the top two spots in the rankings, succeeding their predecessors. This launch also resulted in the ousting of GLM-5.1, developed by the Chinese company Z.ai, from the podium. In April, GLM-5.1 had become the first open-source model to reach the podium of the WebDevArena.

Setbacks for Google and OpenAI

Google faces a setback with the disappearance of Gemini 3.1 Pro from the upper part of the table. Although Google has never dominated this ranking, it consistently held honorable positions alongside OpenAI. The latter, led by Sam Altman, is also experiencing a decline, with GPT-5.5 "High" now sitting in tenth place. This model is surpassed by Muse Spark from Meta, which makes a notable entry in eighth position.

AI Model Rankings for May 2026

Here are the rankings of the top ten models for web development:

Claude Opus 4.7 Thinking (Anthropic): 1,568
Claude Opus 4.7 (Anthropic): 1,562
Claude Opus 4.6 (Anthropic): 1,549
Claude Opus 4.6 (Anthropic): 1,544
GLM-5.1 (Z.ai): 1,532
Claude Sonnet 4.6 (Anthropic): 1,526
Kimi-k2.6 (Moonshot): 1,525
Muse Spark (Meta): 1,509
Claude Opus 4.5 (Anthropic): 1,491
GPT-5.5 High (OpenAI): 1,490

How the WebDevArena Ranking Works

The WebDevArena, like other rankings from LMArena, formerly known as Chatbot Arena, uses a blind duel system. Users of the platform pit two models against each other that have responded to the same request. Each model has an Elo score, which evolves over time through encounters. Defeating a higher-ranked model earns Elo points, while losing to a lower-ranked model penalizes the score. A leading model can continue to progress by winning duels against theoretically weaker opponents, but its score evolves more slowly.