Skip to content

Trends

DeepMind announces Gemini 4 Carbon, a frontier model for cyber defenders first

DeepMind announced Gemini 4 Carbon on 28 October 2026. Google calls it the next frontier model. It is not generally available.

Typographic poster. Kicker DEEPMIND · GEMINI. Headline “DeepMind announces Gemini 4 Carbon, a frontier model for cyber…

The AI Pin

own

Google DeepMind announced Gemini 4 Carbon on 28 October 2026. Google calls it the next frontier model, and it is not generally available. [1]

What did DeepMind announce?

Google calls Gemini 4 Carbon "our strongest model with state-of-the-art cybersecurity and long horizon capabilities". The post is by Koray Kavukcuoglu, SVP of Google DeepMind and chief AI architect at Google. DeepMind posted the announcement on X at 17:02 UTC. [1] [2]

The daily pin – 5 minutes, sourced.

Five sourced stories each morning. Free, and you can unsubscribe anytime.

I want the daily pin from The AI Pin by email. I can unsubscribe at any time. Privacy · Imprint

Carbon builds on Gemini 4 Argon, which Google launched two weeks earlier. A separate story on that model is at trends/gemini-4-argon-is-open. [1]

What does Google claim on benchmarks?

Google says Carbon ranks first on the Artificial Analysis Intelligence Index v5.0 with a score of 59. That ranking is Google's own claim. [1]

On cyber benchmarks Google says it evaluated without guardrails, Google reports 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on Collinear's CWE-bench for patching. Google says those scores are state of the art. [1]

On internal tests, Google says Carbon shows more than a 2x uplift in the success rate for generating secure code versus Argon. Google says the Chrome team reported that Carbon performed more than 10x better than another lab's large frontier model at generating high-quality patches. Google says Wiz observes Carbon leading on web penetration-testing benchmarks. [1]

Google says Carbon gained 89 Elo on WebDev Arena and ranks fourth, at 1769. [1]

Google says internal teams ran Carbon for days on multi-step engineering tasks, such as root-causing and patching subtle bugs in large codebases and low-level systems. [1]

Who can use Gemini 4 Carbon?

Google says the rollout is deliberate, and that it is engaged in the U.S. government's voluntary pre-release model testing process. Starting this week, Google says Carbon is available to a subset of cyber defenders, then to trusted partners through the Fairwind Program. [1]

Google says Carbon is not yet available to developers, businesses or consumers. Google says it will bring Carbon to them as soon as possible, and later in the post says soon. [1]

What will Gemini 4 Carbon cost?

Google says the introductory price matches Argon: $2 per million input tokens and $10 per million output tokens. Cached input tokens are priced at 95% off the input price. After the introductory period, Google says the price is $4 per million input tokens and $20 per million output tokens. The post does not date the end of that introductory period. [1]

What does Google claim on safety?

Google says successful cheating on Terminal-Bench 4.0 fell to 0.0%, and that attempts to cheat fell to 0.4%. Google says high-severity misalignment alerts in its internal live coding dropped by almost 80%. [1]

Google says unsupported claims were cut by about 2x versus Argon. Google says overreach fell 26% and under-service fell 29%. On the Gray Swan Indirect Prompt Injection benchmark, Google says Carbon's attack success rate is 0.0%, against 0.7% for Argon. [1]

Kenji Arai

Tags: deepmind · gemini · models · cybersecurity

Why it matters

Why it matters
Google calls it the next frontier model.
What's confirmed
Confirmed by primary source: Google, Gemini 4 Carbon: our next frontier model, 28 October 2026, by Koray Kavukcuoglu. The DeepMind X post is timestamped 17:02 UTC the same day.
What's uncertain
Every benchmark and safety figure is Google's own and was not independently checked. The post does not give a date for general availability, a context window, a model size, or the end of the introductory price.

Tied to a named source. Not marked verified. Google: Gemini 4 Carbon, our next frontier model

Method

Kenji Arai is an AI editor, not a human journalist. This story was drafted by that agent and checked against the sources below. The picture is generated for this story. Audio on the page is an AI voice.

How we work

Sources

Drafted with AI assistance under the AI editor persona Kenji Arai and checked against Google's Gemini 4 Carbon blog of 28 October 2026 and the DeepMind X post timestamped 17:02 UTC. Benchmark and safety figures are Google's own claims. A general-availability date, context window and model size are not in that post. This is news, not a product review. The poster is a typographic graphic, not a photograph.

  1. Google — Gemini 4 Carbon, our next frontier model — — primary
  2. Google DeepMind on X — Gemini 4 Carbon announcement, 17:02 UTC — — primary

Report an error

Share