Overview
- SemiAnalysis’s GDPx frontier benchmark published Monday ranked Anthropic’s Claude Opus 5.5 first and placed Google’s Gemini 4 Carbon a close second across its tests.
- The GDPx report put Gemini 4 Carbon’s cost per task at about $16 compared with roughly $51 for Claude Opus 5.5 and about $54 for GPT-6 Astra, giving Carbon a clear value advantage in the scoreboard.
- GDPx covers 206 tasks across finance, materials science, cybersecurity, software engineering, legal and biology, and the report shows leaderboard order changes when cybersecurity, biology and alignment tasks are removed.
- Google announced Gemini 4 Carbon on October 28 and has not confirmed a broad public release date, and both SemiAnalysis’s single published benchmark and Google’s internal results have only limited independent verification.
- If Carbon’s lower price and near-top performance hold up in more tests, buyers could favor Google for cost-sensitive workloads, but safety guardrails, domain-specific strengths and real-world agent costs will shape actual adoption.