SemiAnalysis' GDPx puts AI agents to work; Claude Opus 5.5 leads at 12 percent

SemiAnalysis designed GDPx to differ from OpenAI's GDPval by providing a realistic work environment rather than delivering all context in the prompt. The benchmark's 206 tasks span eleven domains, intentionally excluding administration and insurance, because the team argues the first trillion dollars in AI revenue will come from industries where frontier AI delivers the biggest returns. The results show a narrow spread at the top: Claude Opus 5.5 leads at 12 percent, with Gemini 4 Carbon close behind at 10.9 percent but at roughly a third of the cost per task ($16.29). GPT-6 Astra scores 9.6 percent. A filtered analysis reveals that provider safety mechanisms can reshuffle rankings: without the cyber, biology, and alignment domains, GPT-6 Astra edges past Opus 5.5 at 11.8 percent. The secrecy of the tasks, which come from 14 specialized partners, is intended to prevent labs from optimizing for the test, but critics like developer Josh Purtell call GDPx "more noise than signal," citing provider-specific agent software and small sub-domain sample sizes as sources of volatility.