Google announced Gemini 4 Carbon on October 28, 2026, calling it its strongest model yet, with state-of-the-art cybersecurity and long-horizon capabilities. Why it matters: Carbon arrives just two weeks after Gemini 4 Argon, tops a major intelligence index on Google’s reported numbers, and is rolling out first to cyber defenders — not to the general public.
October 28, 2026
What happened
Google DeepMind introduced Gemini 4 Carbon as the next model in its Gemini 4 generation, building on Argon, which launched two weeks ago. Google says researchers and its own models worked together to advance capabilities faster while strengthening safeguards in tandem.
Carbon will launch at the same introductory price as Argon: $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off the input price, according to Google.
The numbers Google is reporting
Google says Carbon ranks #1 on version 5.0 of the Artificial Analysis Intelligence Index, at 59. On cybersecurity, it reports a firm lead across the defense lifecycle: 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. Those external benchmarks were evaluated without guardrails, Google notes.
On internal testing, Google says Carbon delivers more than double Argon’s success rate at generating secure code, and that its Chrome team saw Carbon produce high-quality patches more than 10 times better than another lab’s large frontier model. On web development, Google says Carbon gained 89 Elo points to reach #4 at 1768 on WebDev Arena. All of these figures are company-reported and have not been independently verified.
Built for work that runs for days
Google positions Carbon for sustained, agentic workflows: internally, its teams have seen the model run for days, pushing through dead ends to root-cause and patch subtle bugs across large codebases and low-level systems. The company says it focused improvements on coding, agentic and scientific tasks, full-stack web development, and multimodal understanding, based on feedback after Argon.
On safety, Google says Carbon cut successful cheating on Terminal-Bench 4.0 to 0.0%, reduced high-severity misalignment alerts by almost 80% in live internal coding use, cut unsupported claims by roughly half versus Argon, and scored 0.0% attack success on the Gray Swan indirect prompt-injection benchmark, down from Argon’s 0.7%.
Who gets it, and when
There is no general launch yet. Google says it is taking a deliberate approach, is engaged in the U.S. government’s voluntary process for pre-release model testing, and will make Carbon available starting this week to a subset of cyber defenders, then gradually to additional trusted partners through its Fairwind Program. Wider access for developers, businesses and consumers is promised as soon as possible, after further feedback and guardrail work.
Why it matters
For readers, the practical take is twofold. First, the frontier pace is accelerating: a second Gemini 4 model in two weeks, at unchanged introductory pricing, puts pressure on every rival lab. Second, the defenders-first rollout is becoming Google’s pattern for cyber-capable models — the people who patch software get the tool before everyone else does. If Carbon’s reported patching gains hold up outside Google’s own tests, security teams could close vulnerabilities materially faster. Until independent results land, treat the benchmark lead as a claim to watch, not a settled fact — everything AI, tested, means waiting for the test.
Frequently asked questions
What is Gemini 4 Carbon?
It is Google DeepMind’s newest frontier model, announced October 28, 2026, and described by Google as its strongest model, with a focus on cybersecurity and long-horizon coding and agentic work.
How much does Gemini 4 Carbon cost?
Google says Carbon launches at Argon’s introductory price: $2 per million input tokens and $10 per million output tokens, with cached input at 95% off.
Can I use Gemini 4 Carbon now?
Not yet, unless you are in the first cohort. Google is starting with a subset of cyber defenders this week, expanding through its Fairwind Program, with developer, business and consumer access to follow.
How does Carbon compare with Argon?
Google reports gains across coding, agentic and scientific tasks, more than double the secure-code generation success rate, and stronger alignment results, including zero successful prompt-injection attacks in its reported testing.
Sources: Google (blog.google), 9to5Google