AI Guides
PowerfulAI.tools Editorial Team
0 Comments
By PowerfulAI.tools Editorial Team
Google has announced Gemini 4 Carbon, which it calls its next frontier model and its strongest model so far. It arrives just two weeks after Gemini 4 Argon, and Google is pitching it mainly at long, multi-step coding work and cybersecurity.
There is a catch for everyday users: you can't use it yet. Here is what Google announced on 28 October 2026, who gets access first, what it will cost developers, and what it means for the rest of us.
Gemini 4 Carbon is a new top-end model in Google's Gemini 4 family. Google says it is a step up from Argon on most benchmarks, with gains in coding, agentic tasks (where the AI carries out a series of steps on its own), scientific tasks, full-stack web development and understanding images and other media.
Two areas get the most attention in Google's announcement:
Google also says Carbon ranks first on version 5.0 of the Artificial Analysis Intelligence Index, an independent leaderboard, with a score of 59. These are Google's own results.
For now, access is tightly limited. Google says that starting this week it will make Carbon available to "a subset of cyber defenders", then gradually extend access to more trusted partners through its Fairwind Program.
Google says it is taking "a deliberate approach to rollout". It is collecting feedback and refining safety guardrails before making Carbon available to developers, businesses and consumers through its products. It is also taking part in the US government's voluntary pre-release testing process for AI models.
Google has not given a date for wider release. Its announcement does not name the Gemini app, Google AI Pro or Ultra plans, Google AI Studio or Vertex AI, and it does not list countries. So there is no confirmed date for Australian users or anyone else outside the early-access group.
Google has published developer pricing, charged per million tokens (a token is a small chunk of text). All prices are in US dollars.
The lower rates are a launch offer, not the long-term price. Google has not said how long the introductory period will last.
Because Carbon is strong at finding and exploiting security flaws, Google is putting safety front and centre. It says Carbon uses the same layered safety approach as Argon. That covers how the model is trained, real-time checks for misuse, resistance to prompt injection (hidden instructions planted in web pages or files) and monitoring of the model's reasoning.
Google reports that Carbon had a 0.0% attack success rate on the Gray Swan indirect prompt injection test, down from 0.7% for Argon. It also says Carbon makes about half as many unsupported claims as Argon. Google has published a Gemini 4 safety and security report with more detail.
If you use Gemini today, you won't see Carbon yet, and Google hasn't said when you will.
The bigger story is the direction. Google is releasing new frontier models every few weeks, and it is now holding back its most capable model until security teams have had a head start. That pattern matters for anyone who relies on software, because a model that is good at finding security holes can help defenders fix them, but could also be misused.
For developers and businesses, the published pricing gives an early guide to costs once access opens up. If you are choosing between Gemini models now, Gemini 4 Argon is the current option, and Google says Carbon will launch at the same introductory price.
We checked every claim in this article against Google's own announcement, Gemini 4 Carbon: our next frontier model by Koray Kavukcuoglu (published 28 October 2026), and Google's linked Gemini 4 safety and security report. Benchmark results are as reported by Google. Details we could not confirm on Google's official pages, including a public release date and Gemini app availability, have been left out. Checked on 29 October 2026.
Leave a Comment