Google's new Gemini 4 Carbon AI goes to cyber defenders first
Google says it is its strongest model yet for cybersecurity and long engineering jobs, but most people will have to wait.
Google has announced Gemini 4 Carbon, its newest and strongest AI model. Google says it is built for cybersecurity and for long, hard engineering jobs that take many steps.
Carbon builds on Gemini 4 Argon, which came out two weeks ago. Google reports that Carbon leads on three outside security tests. It scored 88.2% on a test for finding real flaws in software. It scored 54.8% on a test that checks whether a flaw can really be used by an attacker. It scored 71.0% on a test for turning a found flaw into a working fix.
On Google's own tests, Carbon was more than twice as good as Argon at writing safe code. Google's teams have also seen it keep working for days on a hard problem, like hunting down a hidden bug in a huge pile of code.
Most people cannot use Carbon yet. Google says it is rolling it out slowly and is taking part in the U.S. government's voluntary testing of new models before release. Starting this week, a small group of cyber defenders gets it first. More trusted partners will follow through Google's Fairwind Program. Google says developers, businesses and everyday users will get it "as soon as possible."
When it opens up, Carbon will cost the same as Argon. That is $2 for every million tokens it reads and $10 for every million tokens it writes. A token is a small piece of text. Reused text that is saved, or cached, costs 95% less.
Open it and check the claim yourself. That is the point.