SemiAnalysis releases GDPx, an agentic benchmark where the best model scores 12%

SemiAnalysis has released GDPx, a benchmark of agentic capabilities in domains it considers the most valuable, including chip design, cyber and biology. It says even the best model scores just 12%. A post relaying the results reports Claude Opus 5.5 on top at 12.0% and the upcoming Gemini 4 Carbon at 10.9%; that is the relayer's reading, not SemiAnalysis's own text.

• Domain choice: SemiAnalysis argues that the first trillion dollars of AI revenue will be concentrated in sectors that can deploy frontier intelligence at high return, so it rejects benchmarks weighted by share of GDP or by occupation taxonomy. It includes chip design, cyber and bio over domains such as government, real estate and insurance.
• Safety classifiers: SemiAnalysis says Opus 5.5 and GPT-6 Astra have lower overall scores because their safety classifiers sometimes trigger in cyber, biology and alignment tasks. Excluding those three domains gives a different ranking, which it says matches its subjective impression of the models. It calls Muse Wonder Full the best available cyber model today by a meaningful margin.
• How it was built: each domain was made with at least one specialized data partner, with SemiAnalysis doing final quality checks and task iteration. Fireworks provided inference for open-source models and Modal the sandbox infrastructure.

The announcement thread had 262 points on X by November 1, and the results are SemiAnalysis's own, not independently verified.