Google has launched its latest frontier AI model, Gemini 4 Carbon, which delivers improvements in coding, agentic tasks, and cybersecurity defense over its previous Argon model released earlier this month. The new model currently takes the top spot on the Artificial Analysis Intelligence Index (v5.0) benchmark, outperforming rival models including Claude Opus 5.5 and GPT-6 Astra.

According to Google, Gemini 4 Carbon demonstrates a significant leap in cyberdefense capabilities, excelling at finding vulnerabilities, resisting prompt injection attacks, and generating secure code. Internal teams noted the model can run continuously for days to handle complex, multi-step engineering assignments and reduce unsupported claims relative to Argon.

Gemini 4 Carbon is initially available to a select group of cybersecurity defense companies before expanding to partners in Google’s Fairwind Program. The model launches at an introductory price of $2 per million input tokens and $10 per million output tokens.

Why it matters

  • Gemini 4 Carbon demonstrates advanced multi-day autonomous execution for complex engineering and bug-patching workflows.

  • Initial access is restricted to select security partners and the Fairwind Program before a broader enterprise rollout.

  • Competitive pricing at $2/$10 per million tokens sets an aggressive baseline for high-reasoning frontier models.

Source: theverge.com