Claude Fable 5.2 tops SemiAnalysis's GDPx at 16.4%, ahead of Opus 5.5's 12.0%, and is the priciest model on the board at $88 a task

Claude Fable 5.2 is first on GDPx, SemiAnalysis's benchmark of frontier models on high-value professional work, with a score of 16.4%. That is 4.4 points ahead of Claude Opus 5.5, the previous leader at 12.0%, and almost double the 8.5% of Fable 5.1, the model it replaces. ArtificialWatch read the figures from the GDPx leaderboard at 01:50 UTC on November 10.

THE TOP FIVE

From SemiAnalysis's leaderboard, all domains:

• Claude Fable 5.2: 16.4%, $88.15 per task
• Claude Opus 5.5: 12.0%, $51.41 per task
• Gemini 4 Carbon: 10.9%, $16.29 per task
• GPT-6 Astra: 9.6%, $53.99 per task
• Muse Wonder Full: 9.0%, no price listed

WHAT GDPX MEASURES

GDPx v1.0 has 206 tasks across 11 domains, built with 14 specialist data partners: chip design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement, hardware design, materials science, alignment and software engineering. The score weights every domain equally and averages three attempts per task. Before Fable 5.2, no model on the leaderboard reached 13%.

THE COST

Fable 5.2 is also the most expensive model on the board to run. Its $88.15 per task is 71% more than Opus 5.5's $51.41, and it wrote 1.23 million output tokens per task, against 548,000 for Opus 5.5. The cheapest models score far lower: GPT-6 Luna and DeepSeek 4.1 Flash 1017 cost under $1 per task and score 5.1% and 5.4%.

WHAT WE DO NOT KNOW

Which effort setting SemiAnalysis ran, and how Fable 5.2 splits across the 11 domains, since the leaderboard's per-domain view loads in the browser and we did not read it. @scaling01 on X pointed out the new ranking; the figures above come from SemiAnalysis's own leaderboard.