What Is Gemini 4 Carbon, and Can You Use It Yet? (October 2026)
Not yet. Google announced Gemini 4 Carbon, its new top model, on October 28, 2026, but it is going only to a small group of vetted cyber defenders first. It isn't in the Gemini app or on any Google AI plan, and Google hasn't given a date for developers, businesses, or consumers. Here's what Carbon is, how it compares with Argon, and what to do now.
See Claude set up for your job
Skip the theory — pick your profession and get the real workflows, ready-to-use prompts, and exact setup for your work.
TL;DR: Gemini 4 Carbon is Google's newest and highest-scoring AI model, announced October 28, 2026. You can't use it yet unless you're one of the vetted cyber defenders Google is giving it to first. It is not in the Gemini app or on any Google AI plan, and Google hasn't said when it will reach developers, businesses, or consumers. Google's best model you can use today is still Gemini 4 Argon, included with Google AI Pro ($19.99/month) and Ultra.
If you're wondering whether to change anything this week: you don't need to. Carbon is a preview of what's coming to Gemini, not something you can pick in the model menu. This post covers what Carbon is, how it compares with Argon and the competition, and what to watch for.
What is Gemini 4 Carbon?
Carbon is the second model in Google's Gemini 4 generation. Google's announcement, written by Koray Kavukcuoglu (senior vice president of Google DeepMind and Google's chief AI architect), says it builds on Gemini 4 Argon, which launched two weeks earlier, and is designed for "much harder, long-horizon engineering problems."
Google's headline claims:
- Top of the main independent ranking. Carbon is ranked first of 52 models on the Artificial Analysis Intelligence Index (version 5.0), with a score of 59. Artificial Analysis lists the same result on its own site.
- Cybersecurity. 88.2% on SEC-bench Pro (finding real vulnerabilities), 54.8% on ExploitGym (confirming a flaw can be exploited), and 71.0% on CWE-bench (fixing the root cause). Google notes these outside benchmarks were run without its safety guardrails switched on.
- Fewer made-up claims. On Google's internal tests, Carbon makes about half as many unsupported claims as Argon and produces secure code more than twice as often.
- Prompt-injection resistance. 0.0% attack success on the Gray Swan indirect prompt-injection benchmark, down from Argon's 0.7%. The Decoder points out that this doesn't mean prompt injection is solved.
Most of these are Google's own numbers. The independent checks so far are below.
Can you use Gemini 4 Carbon now?
Almost certainly not. As of October 28, 2026:
| Who | Carbon access |
|---|---|
| Vetted cyber defenders (selected by Google) | Starting this week |
| More "trusted partners" via Google's Fairwind Program | Gradually, no dates given |
| Google AI Pro / Ultra subscribers | No; plans list Argon, not Carbon |
| Free Gemini and Google AI Plus | No |
| Developers on the Gemini API | Not yet; Carbon isn't on the API pricing page |
| Google Workspace / Gemini Enterprise | Not announced |
Google's announcement says: "Starting this week, we will make Carbon available to a subset of cyber defenders," with more partners added through the Fairwind Program "before making Carbon available to developers, businesses, and consumers," which Google says it aims to do "as soon as possible." It gives no date.
When we checked on October 28, Google's subscriptions page still listed Gemini 4 Argon on AI Pro and Ultra and did not mention Carbon. Google also hasn't said which plans will get Carbon when it does open up, so don't assume it will land on the same plans as Argon.
For what it's worth, Argon took two weeks to go from cyber defenders only (September 30) to paying subscribers (October 13). Google hasn't committed to the same timeline for Carbon.
How does Carbon compare with Argon, ChatGPT, and Claude?
Two independent benchmark groups have already published results:
- Artificial Analysis ranks Carbon first of 52 models on its Intelligence Index, a composite of reasoning, knowledge, and coding tests.
- Vals AI, which tests models on finance, legal, tax, and coding work, ranks Carbon second of 29 on its Vals Index at 70.36%. That's just behind Claude Sonnet 5.5 (70.85%) and just ahead of Claude Opus 5.5 (70.18%). Carbon is first on the Vals Finance Index and on Legal Research Bench.
According to The Decoder, Carbon still trails Claude Opus 5.5 on Artificial Analysis's coding-agent measure, and it uses noticeably more "thinking" tokens than Argon, so some finance and tax tasks took more than twice as long.
The practical reading: Carbon is a real step up from Argon and is in the same top group as Anthropic's and OpenAI's best models, without clearly beating them at everything. For professional work like legal research and financial analysis, the early independent numbers are strong. None of that matters until you can use it.
How much will it cost?
In the Gemini app: unknown. Google hasn't said which subscriptions will include Carbon.
For developers: Google's announcement gives an introductory price matching Argon's, $2 per million input tokens and $10 per million output tokens, with cached input 95% cheaper, rising to $4 / $20 after the introductory period. Artificial Analysis already lists Carbon at the $4 / $20 rate. Carbon is not yet on Google's Gemini API pricing page, and Google hasn't said on its own pages when the introductory price ends. Check the pricing page once Carbon appears there, and budget at $4 / $20.
Why is Google holding it back?
Google says Carbon is state of the art across the whole cybersecurity "defense lifecycle": finding vulnerabilities, proving they can be exploited, and patching them. Those same skills could help attackers. So Google is releasing it to defenders first and says it is "actively engaged in the U.S. government's voluntary process for pre-release model testing." Google says Carbon uses the same layered safety system as Argon.
This "defenders first, everyone else later" pattern is now standard for the most capable models. Google used it for Argon, and Anthropic keeps its most cyber-capable model, Claude Mythos, in a restricted program.
Should you change anything now?
- If you pay for Google AI Pro or Ultra: nothing changes this week. Argon is still the best model in your menu. Keep testing it on real work so you have a baseline to compare Carbon with later.
- If you're on free or AI Plus: don't upgrade for Carbon. No plan includes it yet, and Google hasn't said which will.
- If you build on the Gemini API: you can't switch to Carbon yet. When it appears, test it on your own tasks and plan costs at $4 / $20, not the launch price.
- If you work in security: Google's Fairwind Program is the route to early access. Google hasn't published how organizations are chosen.
- If you're choosing between tools: Carbon's arrival doesn't change today's choice. Our ChatGPT vs Claude vs Gemini guide covers what you can actually use now.
We'll update this post when Google says who gets Carbon and when.
Sources
- Google, Introducing Gemini 4 Carbon (Koray Kavukcuoglu, October 28, 2026)
- Google, Google AI plans and pricing (checked October 28, 2026)
- Google, Gemini API pricing (checked October 28, 2026)
- Artificial Analysis, Gemini 4 Carbon: Intelligence, performance and price analysis (checked October 28, 2026)
- Vals AI, Gemini 4 Carbon (checked October 28, 2026)
- The Decoder, Google Deepmind releases Gemini 4 Carbon just two weeks after Argon with gains across the board (October 28, 2026)
See Claude set up for your job
Skip the theory — pick your profession and get the real workflows, ready-to-use prompts, and exact setup for your work.
Set up AI for your job — free, in about 2 minutes
Pick your profession and get your first working AI tool, a step-by-step guide, and a $0 plugin to take home. No credit card.
Get my free setupSee Claude set up for your job
Real workflows and ready-to-use prompts, profession by profession.
Frequently asked questions
What is Gemini 4 Carbon?+
Gemini 4 Carbon is Google's newest frontier AI model, announced on October 28, 2026 by Koray Kavukcuoglu, Google's chief AI architect. It builds on Gemini 4 Argon, which launched two weeks earlier. Google says Carbon is state of the art in cybersecurity and in long, multi-step work such as large software projects.
Can I use Gemini 4 Carbon in the Gemini app?+
No, not as of October 28, 2026. Google is giving Carbon first to a subset of cyber defenders, then to more trusted partners through its Fairwind Program. Google's subscriptions page still lists Gemini 4 Argon, not Carbon, on AI Pro and Ultra. Google says it will make Carbon available to developers, businesses, and consumers as soon as possible, but has not given a date.
Is Gemini 4 Carbon available on the Gemini API?+
Not for general developers yet. Google's announcement gives introductory API pricing of $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after the introductory period, but Carbon is not yet listed on Google's Gemini API pricing page. Check that page for the end date of the introductory price before budgeting.
Is Carbon better than Argon?+
On Google's own tests, yes, and independent rankings largely agree. Artificial Analysis ranks Carbon first of 52 models on its Intelligence Index with a score of 59. Vals AI ranks it second on its professional-work index, just behind Claude Sonnet 5.5 and ahead of Claude Opus 5.5, and first on its Finance Index and Legal Research Bench. Google also says Carbon makes about half as many unsupported claims as Argon on its internal tests.
Why is Google restricting Carbon?+
Google says Carbon is very good at finding, exploiting, and patching software vulnerabilities, skills that could also help attackers. It is releasing Carbon to defenders first and taking part in the U.S. government's voluntary pre-release testing process. Google followed the same path with Argon, which went to cyber defenders on September 30 and reached AI Pro and Ultra subscribers on October 13.
Should I upgrade my Google AI plan to get Carbon?+
No. No Google AI plan includes Carbon yet, and Google hasn't said which plans will get it. If you need Google's best model available today, that's Gemini 4 Argon, which comes with Google AI Pro ($19.99 per month in the US) and Ultra.
Related Guides
Can Gemini Edit Documents and PDFs Right Inside the Chat?
Yes, as of October 29, 2026. Gemini can now show a live preview of a document, PDF, spreadsheet, slide deck, or code file it creates, and you can edit that file in the chat by typing directly or by highlighting a part and asking for a change. It's free for all Gemini users. On the web it covers Google Docs, Sheets, slides, Markdown, TXT, code, and PDFs; on mobile it's PDFs only. Here's how it works and what to check before you rely on it.
What Is Claude Duet, and Should You Join the Waitlist? (October 2026)
Anthropic opened a research preview of Claude Duet on October 21, 2026: a separate app where Claude works as a proactive agent, sending a morning brief, watching your connected apps, and asking before it sends, pays, signs, or deletes anything. Here's what it does, who can get it, what it costs, and how to try it safely.
How Do I Lock a Chat in ChatGPT? What Locked Chats Hide — and What They Don't
OpenAI began rolling out Locked Chats on October 20, 2026. You can put individual ChatGPT conversations behind a PIN or Face ID/fingerprint so they drop out of your history and search. Here's how to turn it on, what it protects, and the PIN-reset catch that permanently deletes everything.