Technorat Tech news and explainers, minus the noise · No ads · No tracking · Sources on every post
Technorat
Search

News/AI

Google releases Gemini 4 Carbon: what changes

Google released Gemini 4 Carbon on November 10, 2026. See who can use it, what it costs per million tokens, and how it differs from Gemini 4 Argon.

Short answer

Google made Gemini 4 Carbon, its strongest model for long agentic tasks and cybersecurity work, available on November 10, 2026 to paid API users, Google AI Ultra subscribers and business customers. Introductory API pricing is $2 per million input tokens and $10 per million output tokens until November 30, 2026.

Key takeaways

  • Gemini 4 Carbon sits above Gemini 4 Argon and targets hard multi-step coding, research and security tasks.
  • API pricing doubles after November 30, 2026, to $4 per million input tokens and $20 per million output tokens.
  • Google says Argon stays the recommended default for most everyday engineering and knowledge work because Carbon uses more tokens.

Google released Gemini 4 Carbon on November 10, 2026, in a post by Koray Kavukcuoglu of Google DeepMind. The company calls Carbon its strongest model for cybersecurity and long-horizon agentic work, and it now reaches paid developers, Google AI Ultra subscribers and enterprise customers. Here is what Google announced, what it costs and what the announcement leaves open, as of November 2026.

What did Google announce?

Google announced Gemini 4 Carbon, a model positioned above Gemini 4 Argon for difficult multi-step coding, debugging, research and security tasks.

Carbon is a large language model (if the term is new, see our explainer on what a large language model is). Google says it uses more tokens than Argon to explore and verify its work, which is why Argon remains the recommended choice for most day-to-day engineering and knowledge work.

The model first went to cybersecurity defenders through Google's Fairwind Program about two weeks before the public release. Alongside the launch, Google previewed a /swarm feature in Google Antigravity for coordinating teams of agents.

Google's own claims include:

  • First place on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench, at what Google describes as more than 2x lower cost per task than the next best external model.
  • Second place on the new Multimodal Knowledge Agent (MKA) benchmark.
  • Leading results on SEC-bench Pro, ExploitGym and CWE-bench v1, covering finding, exploiting and patching flaws.

These figures come from Google and its partners. Some, such as the Project Fortify results, rest on internal benchmarks that are not fully published.

Who can use it and when?

Carbon became available on November 10, 2026, through the paid Gemini API, Google AI Ultra plans and Google's business products, with rollout continuing in some apps.

Group Where Carbon appears
Developers (paid tier) Gemini API via Google AI Studio and Android Studio
Enterprises Gemini for business, Google Antigravity, CodeMender
Google AI Ultra subscribers Gemini app, AI Mode in Search, AI Studio, Antigravity (rolling out)
Scientists Free six-month access by application, subject to selection
Cyber defenders Fairwind Program, by application

Google's post does not give country-level availability, so check your account before planning around it.

What does it cost?

Introductory API pricing matches Argon at $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after November 30, 2026.

That doubling matters for teams estimating budgets. A workload that spends heavily on output tokens, such as long agent runs that write and rewrite code, will see the biggest change on December 1. Because Carbon also consumes more tokens per task than Argon, the per-token price is only part of the cost picture.

What changes for safety and security?

Google reports lower attack success rates than Argon in its red-teaming, but says resistance to prompt injection still needs work.

According to the announcement, Carbon blocked 99% of attack trajectories from an adaptive red-teaming agent in simulated environments, with a per-attempt breach rate of 0.8% versus 3.1% for Argon. On a 375-task agentic honesty benchmark it passed 92% of tasks, against 70% for Argon.

Google also writes that models "could be further improved against sophisticated attacks in some distributions." Agents that read web pages or email remain exposed to injected instructions, and like any model Carbon can still state wrong things with confidence (our guide to AI hallucinations covers why).

What is still unclear?

Independent benchmarks, regional availability and the developer-facing limits are not spelled out yet.

The developer post does not list a context window, model ID or rate limits, and it does not mention Vertex AI. Broader scientific benchmarking is still underway, according to Google. If you plan to give Carbon access to work files or mail through the Gemini app, review the privacy settings for AI chatbots first, then test it on a contained task before moving real workloads over.

Frequently asked questions

Can free Gemini users try Gemini 4 Carbon?

Google's announcement lists paid-tier API users, Google AI Ultra subscribers and business customers. Ultra subscribers get it in the Gemini app, AI Mode in Search, Google AI Studio and Google Antigravity. A free tier is not mentioned in the post.

How much does Gemini 4 Carbon cost in the API?

Google lists an introductory price of $2 per million input tokens and $10 per million output tokens, the same as Argon. After November 30, 2026, the price for the public Carbon model is $4 per million input tokens and $20 per million output tokens.

What is the Fairwind Program?

It is Google's access program for cybersecurity defenders. Carbon reached Fairwind participants about two weeks before the public release, and its cyber capabilities remain gated through applications to the program.

Should I switch from Gemini 4 Argon to Carbon?

Google itself recommends Argon for most software engineering and knowledge work. Carbon spends more tokens exploring and verifying, so it fits long, complex tasks where that extra work pays off.

Sources

  1. Google: Gemini 4 Carbon is now available
  2. Google: Gemini 4 Carbon is now available for developers

Research-based: Technorat does not claim hands-on testing. Facts as stated by the sources above, as of November 2026. See how we research. Spotted an error? contact@technorat.com

#AI#LLMs#Google