SemiAnalysis launches GDPx AI-agent benchmark; top model scores 12.0% on its leaderboard

SemiAnalysis launched GDPx on Nov. 2, a benchmark for AI agents in domains it considers especially valuable. The GDPx v1.0 site lists 11 domains, 206 tasks and 14 data partners.

The domains are chip design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement, hardware design, materials science, alignment and software engineering. SemiAnalysis chose them over industries such as government, real estate and insurance. Its rationale is a forecast, not a measured economic result: it expects early AI revenue to concentrate in sectors that can put frontier models to work at high returns. The company says it worked with at least one specialized data partner in each domain, was responsible for final quality assurance and was heavily involved in task iteration.

On the , Claude Opus 5.5 leads at 12.0%, followed by Gemini 4 Carbon at 10.9% and GPT-6 Astra at 9.6%. GDPx weights its 11 domains equally and weights partners equally within each domain. Tasks receive binary or continuous grades, depending on the domain; the default score averages three attempts. That 12.0% is a result under GDPx’s scoring rules, not a measure of how often the model succeeds at work generally.