Google Develops Gemini 4 Carbon to Challenge Rivals as Model Race Intensifies

Google is internally testing an advanced AI model codenamed Gemini 4 Carbon that appears to deliver superior coding performance compared to its recently released Gemini 4 Argon, escalating competition with Anthropic and OpenAI as frontier AI models evolve. The company announced Gemini 4 Argon on September 30, 2026, positioning it as a flagship model for enterprise knowledge work, long-context reasoning, and cybersecurity applications. However, internal employee testing of Carbon—revealed through Business Insider reporting in October 2026—suggests Google is advancing its capabilities even further, with one tester characterizing Carbon's coding performance as comparable to Anthropic's Claude Opus 5.5, a leading model in the market.