SemiAnalysis introduced GDPx, a new benchmark evaluating the general agentic capabilities of frontier AI models across 11 domains, including chip design, biology, and finance. The top model on the initial leaderboard, Anthropic’s Claude Opus 5.5, achieved a 12.0% overall success rate, ahead of Google’s Gemini 4 Carbon at 10.9% and OpenAI’s GPT-6 Astra at 9.6%. No model led every category: Gemini 4 Carbon ranked first in biology and software engineering, while GPT-6 Astra topped the recursive self-improvement and materials science categories. SemiAnalysis plans to update GDPx every three to six months and says the results show advancing capabilities, alongside major barriers to broad proficiency in professional AI tasks.
Source: Read the original article
