Google releases Gemini 4 Carbon, first restricted to cyber defenders
Access is limited for now, with broader availability to follow once guardrails are refined.
Google has released Gemini 4 Carbon, which it positions as its strongest frontier model, scoring 59 on the Artificial Analysis Intelligence Index v5.0 and placing first ahead of Claude Opus 5.5 and GPT-6 Astra. The company says its gains are concentrated in cybersecurity, with 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. Access is initially restricted to a small set of cyber defenders, with wider release to trusted partners through the Fairwind Program and general availability to follow once guardrails are refined. Carbon will carry the same introductory pricing as Argon, priced at $2 for one million input tokens and $10 for one million output tokens, with cached input tokens discounted by 95 percent. Developers should note Google's emphasis on long-running agentic work and its claim that Carbon roughly halves unsupported claims compared with Argon.
Why it matters
For most people who follow technology, the main change is that Google's strongest model is not yet open to the public. Its headline gains are in cybersecurity, where it scores highest on finding vulnerabilities, exploiting them and patching them, so early use is concentrated among defenders. Wider access will depend on Google refining its guardrails and extending the Fairwind Program to more trusted partners. Until then, the model's claimed lead on the Intelligence Index is a company-reported result rather than something most users can test themselves.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.