SemiAnalysis released GDPx on November 2, 2026, a 206-task agentic benchmark spanning eleven high-value domains — and no frontier model cleared even a fifth of it1,3. Anthropic's Claude Opus 5.5 leads the first ranking at 12.0%, followed by Google's Gemini 4 Carbon at 10.9% and OpenAI's GPT-6 Astra at 9.6%.
What GDPx measures
The benchmark, which SemiAnalysis describes as the successor to GDPval, covers chip design, cybersecurity, biology, finance, legal, quantitative trading, recursive self-improvement, hardware design, materials science, alignment, and software engineering. It intentionally excludes sectors such as government, real estate, and insurance, on the thesis that the first trillion dollars of AI revenue will concentrate in domains where frontier intelligence delivers the highest return on investment. Each task carries an eight-hour default time limit, and the headline score reports pass@1 across three attempts.
SemiAnalysis partnered with 14 specialized data partners to build the task set. The GDPx announcement carries the names of Max Kan, Andrew Megalaa, and Dylan Patel, who founded SemiAnalysis.
Domain-level standouts
Claude Opus 5.5 ran in Claude Code, GPT-6 Astra ran in Codex, and Gemini 4 Carbon ran in AntiGravity. Claude Opus 5.5 is the top all-around model and leads in quantitative trading, hardware design, and chip design. GPT-6 Astra wins in recursive self-improvement and materials science. Gemini 4 Carbon is the strongest biology model and the only model in the same tier as Claude Opus 5.5 and GPT-6 Astra. In cybersecurity, Muse Wonder is the leading available model by a meaningful margin.
SemiAnalysis notes that Claude Opus 5.5 and GPT-6 Astra have depressed overall scores because their safety classifiers intervene in cybersecurity, biology, and alignment tasks; excluding those three domains shifts the ranking. In finance, all models score 0% under binary grading, so GDPx reports an average partial score instead.
Competitive picture
OpenAI and Anthropic generally dominate the top of the GDPx ranking, with Gemini 4 Carbon as the closest competitor. Thirteen models were tested in total.
ANALYSIS The 12% ceiling and the 0% binary-grading result in finance underscore how far current frontier models remain from reliable autonomous performance in the domains SemiAnalysis considers most commercially consequential.