Claude Opus 5.5 and Gemini 4 Carbon are the top two models on SemiAnalysis's new GDPx benchmark, with Opus 5.5 scoring 12.0% and Carbon 10.9%, a gap of a little over one point. Carbon reached its score at $16.29 a task, while Opus 5.5 needed $51.41. Opus 5.5 has been generally available since September 22, 2026, while Carbon was announced on October 28 and is still restricted to cyber-defence partners through Google's Fairwind program, with no public model ID. GDPx v1.0, released this week, is the first independent evaluation to score the two models on the same tasks with the same harness and report each run's cost. Claude Opus 5.5 is Anthropic's current flagship for long-running agentic coding and knowledge work, priced at $4 per million input tokens and $20 per million output, with batch requests 50% off at $2 / $10, a 1M-token context window, and adaptive thinking that is always on. Gemini 4 Carbon is Google's newest frontier model, announced by Koray Kavukcuoglu as its strongest model for cybersecurity, long-horizon agents and complex software engineering, priced at $2 per million input tokens and $10 per million output during an introductory period, with the standard rate later $4 / $20