Google announces Gemini 4 Carbon

Google's post on 28 October 2026 announces Gemini 4 Carbon, which it calls its strongest model, aimed at cybersecurity and long-horizon work. Google says Carbon ranks first on Artificial Analysis Intelligence Index v5.0 with a score of 59. On cybersecurity benchmarks evaluated without guardrails, Google reports 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on Collinear's CWE-bench. On internal benchmarks, Google says Carbon more than doubles Argon's success rate at generating secure code. Google also says Carbon gained 89 Elo on WebDev Arena, to 1768, which it places fourth.

The same post says Google is in the U.S. government's voluntary pre-release testing process. This week Carbon goes to a subset of cyber defenders, then to more trusted partners through the Fairwind Program. Google says it is still iterating on guardrails before Carbon is available to developers, businesses, and consumers. The announced introductory API price matches Argon: $2 per million input tokens and $10 per million output tokens, with cached input at 95% off. After that introductory period, Google says the price is $4 per million input tokens and $20 per million output tokens.

Google describes internal coding runs that last for days on multi-step engineering work. On safety, the post says Carbon's successful cheating on Terminal-Bench 4.0 fell to 0.0%, with cheat attempts at 0.4%, and that it scored 0.0% attack success on Gray Swan's Indirect Prompt Injection benchmark. Those figures are Google's.

Who it's for: a cyber defender in the first cohort, or someone waiting until Carbon is on the API and in products.

Fits: Best for Writing. Gemini stays third, for a draft next to Gmail and Docs. Pro and Ultra subscribers can still use Gemini 4 Argon there. Carbon is not on those surfaces yet. Best for Research stays Perplexity. The Gemini alternatives, the Gemini start guide, and the research guide are unchanged. The verdict is unchanged.