WebDev Arena: Fable 5.1 Leads, Qwen Dominates Fullstack

Key Takeaways

1Claude Fable 5.1 Max takes the lead in the overall ranking and front-end with 1,765 points

2Qwen 3.8 Max dominates in fullstack, while Fable 5.1 Max drops out of the top 10 in this category

3The ranking is based on anonymized duels and an Elo score, with limits related to the voting format

Why it matters : This ranking influences the choice of AI models for web development, but it does not replace a thorough evaluation in real-world contexts.

The WebDev Arena rankings for September 2026 are a game changer. Claude Fable 5.1 Max, launched on September 1, takes the top spot overall and in the front-end category with a score of 1,765, while Qwen dominates in the fullstack category. This ranking is based on anonymized duels and an Elo score, with acknowledged limitations for a tool-assisted but not definitive selection.

A Ranking Based on Anonymized Duels and an Elo Score with Safeguards

The WebDev Arena differentiates models through anonymized duels: two systems receive the same prompt, and users vote without knowing their identities. These votes contribute to an Elo score, where a victory against a higher-ranked opponent yields more points. The analysis is limited to the response produced for a single query, without considering code maintainability or its integration into an existing project. A rank spread column indicates a range of ranks, typically broad for newcomers or when scores are not yet stabilized. According to the organizers, these rankings represent a pool of models to experiment with, rather than a definitive decision for team selection.

Fullstack: Qwen 3.8 Max Takes the Lead, Fable 5.1 Out of the Top 10

The ranking now distinguishes between two categories: front-end and fullstack. In the fullstack category, the hierarchy shifts: Claude Fable 5.1 Max drops out of the top 10. Qwen 3.8 Max claims the top position, followed by Claude Opus 5 Max, Kimi K3 Max, then Claude Opus 5 High and GLM 5.3 Max. The previous version from Anthropic, Claude Fable 5, holds the seventh position. DeepSeek enters the top 10 with two models ranked ninth and tenth.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Overall Top 10: 1,765 Points for Fable 5.1 Max and Strong Chinese Presence

Overall, Claude Fable 5.1 Max leads the ranking with a total of 1,765 points. This model, available since September 1, benefits from reduced cyber restrictions and a lower fee for cache access. The gap between it and the next model exceeds 75 points, which is rare for this ranking. Half of the top 10 seats are occupied by Chinese labs. Following closely, Qwen 3.8-max-0902 (Alibaba) is in second place with 1,688 points, ahead of Claude Opus 5 Max, which drops to third after leading the previous month. Next are Kimi K3 Max (Moonshot) at 1,674, Qwen 3.8 Max (Alibaba) at 1,669, and Claude Opus 5 High at 1,661. Grok 4.6 High (SpaceXAI) makes a leap to seventh place with 1,629, an improvement over Grok 4.5. Claude Fable 5 takes the eighth spot with 1,628, followed by Hy4-preview (Tencent) at 1,626 and Qwen 3.8 Flash Next (Alibaba) at 1,622.

Front-end and Technologies: Fable 5.1 Max Maintains the Advantage, Including in React and HTML

In the front-end category, Claude Fable 5.1 Max maintains a clear lead. It surpasses Claude Opus 5 Max and Qwen 3.8-max-0902. The breakdown by technology confirms this position, with leadership also observed in React and HTML.