Google's Gemini 4 Carbon goes to cyber defenders first

Google announced Gemini 4 Carbon on October 28, 2026, calling it "our strongest model with state-of-the-art cybersecurity and long horizon capabilities." Most people cannot use it. Google says: "Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program". Google says Carbon "will launch at the same introductory price" as Gemini 4 Argon, "$2 per million input tokens and $10 per million output tokens". Ars Technica reported that Google declined to say when that price ends.

#What changed?

Google says Carbon follows Gemini 4 Argon, which it launched two weeks earlier. Google says Carbon is "built for sustained agentic workflows" and that "Internally, our teams have seen Carbon run for days" on multi-step engineering tasks such as "root-causing and patching subtle bugs across large codebases and low-level systems."

Google also says it improved "coding, agentic, and scientific tasks, as well as full-stack web development and multimodal understanding."

Ars Technica was more skeptical: "Gemini 4 Carbon is essentially a slightly better version of Argon." Ars added: "for all we know, it could be a post-training upgrade of Argon."

#Who can use it?

• First: "a subset of cyber defenders", in Google's words. Google names none of them.
• Next: "additional trusted partners through our Fairwind Program", phased in gradually.
• Later: Google says it is "continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible." It gives no date.

Google also says it is "actively engaged in the U.S. government’s voluntary process for pre-release model testing." The Verge reported that Carbon is "currently only available to a select group of cybersecurity defense companies". Ars Technica wrote that "Google does not make any promises about the timeline for a wide release".

#The Fairwind Program

Google's Fairwind page says the program "gives high-priority defenders (like governments, healthcare providers, and telecommunications services) early access to advanced models". It says "We currently work with over 650 partners globally" and that defenders can request access through a form. Partners "are only permitted to carry out dual-use tasks, including authorized threat simulation, reverse engineering, and malware analysis for defensive and academic research purposes."

When our sources checked it, the Fairwind page described access to Gemini 4 Argon and did not mention Carbon.

#How much does it cost?

Google's pricing, in its own words:

• At launch: "$2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price."
• Later: "After the introductory period expires, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply."

Google gives no end date for Carbon's introductory period. Ars Technica reported: "Google has declined to provide an end date when asked, suggesting it hasn’t decided when the price will increase." Only Google's first cohort can use Carbon at any price as of October 28, 2026.

#What do the benchmarks show?

Every number below is Google's. Google says the cyber scores were "evaluated without guardrails", and its methodology says "All the results for non-Gemini models are sourced from providers' self-reported numbers unless otherwise mentioned below."

Google says Carbon reaches these scores (from Google's Gemini models page):

Google also says "Carbon jumps +89 elo points to land at #4 (1768) on WebDev Arena". It says "On internal benchmarks, Carbon achieves more than 2x uplift in success rate for generating secure code over Argon," and that its Chrome team reported Carbon "performed more than 10X better than another lab's large frontier model when generating high quality patches." Google does not name that model.

Ars Technica wrote: "The gap between Carbon and the competition is just a few points."

#What does Google say about safety?

Google says Carbon "builds on the same multi-layered safety and security stack of Argon", described in Google's Gemini 4 safety post. Google's safety claims for Carbon:

• "Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0%, with attempts to cheat dropping to 0.4%".
• "high-severity misalignment alerts dropped by almost 80%" in live coding environments with internal teams.
• "Carbon cuts unsupported claims by approximately 2x over Argon."
• On Gray Swan's Indirect Prompt Injection benchmark, Argon had "a very low 0.7% attack success rate" and "Carbon achieves 0.0%".

#How does it compare?

The Verge reported that Google highlights Carbon's top spot on the Artificial Analysis index "with Claude Opus 5.5 and GPT-6 Astra trailing behind." For those rivals, see our coverage of GPT-6 Astra and GPT-6.1 Terra. For other Google DeepMind news in October 2026, see Genie 4.

Our analysis: Carbon's pitch rests on cyber benchmarks that Google ran or reported itself, with rival scores taken from those rivals' own reports. Until outside teams can test it, the margins over Argon and competitors, which Ars Technica called a few points, are Google's claims. The price is clear for launch, but with no end date for the introductory rate, buyers cannot plan on $2 and $10 for long.

#What is still unclear?

• Who is in the first cohort. Google names no organizations.
• Fairwind and Carbon. The Fairwind page named only Argon when our sources checked it.
• When the introductory price ends. Google declined to say, per Ars Technica.
• Wider access. No date for the Gemini API, businesses or consumers.
• Docs. No Carbon model card, safety report or API model ID found.