Arena added GPT-6 Astra Max to its web-development leaderboard on September 5. Its score of 1,797 exceeds Claude Fable 5.1 Max’s 1,762, although both have rank intervals of 1–2. The comparison reflects user preferences between generated frontend applications in this particular evaluation. Astra Max has 1,199 votes and Fable Max has 2,275.
Arena’s September 5 changelog lists the addition of GPT-6 Astra Max to Code Arena: WebDev. The table for that date gives it a score of 1,797 from 1,199 votes. The addition places Astra Max in an evaluation where users choose between generated frontend applications. Claude Fable 5.1 Max appears in the same table with a score of 1,762.[1], [2]
Arena uses paired comparisons of applications generated for real frontend tasks, with user preferences feeding an Elo-style rating system. The leaderboard described by Crypto Briefing covers 126 models and more than 650,000 votes. Those totals describe the broader table, rather than the number of votes cast for Astra Max alone. The evaluation concerns the interfaces generated for these web-development tasks.[2]
Claude Fable 5.1 Max’s score of 1,762 rests on 2,275 votes. Although GPT-6 Astra Max has the higher point estimate at 1,797, Arena gives both models rank intervals of 1–2. That shared interval limits any claim of a clear first-place finish. The result concerns users’ frontend preferences and does not establish a general lead across every coding task.[1], [2]
