Google opened Gemini 4 Carbon, its most powerful AI model, to paying developers and businesses on Tuesday, November 10. The model had been locked behind a limited cybersecurity program since late October. Now anyone on the paid Gemini API tier can call it.
The price is $2 per million input tokens and $10 per million output tokens through November 30. On December 1, both rates double, to $4 and $20. Google AI Ultra subscribers also get Carbon in the Gemini app, AI Mode in Search, and the Antigravity coding platform.
Google calls Carbon its strongest model yet. It is built for long, multi-step work: complex coding jobs, debugging across big codebases, and open-ended scientific research. The model runs extended verification loops and spends more thinking tokens than its sibling, Gemini 4 Argon, to get there.
“Argon remains the ideal choice for the majority of software engineering and knowledge work, while Carbon is built for tasks that test the limits of a model’s intelligence,” Google said in its announcement. The message is plain enough. Argon is the everyday model. Carbon is the one you reach for when the job is hard enough to justify the extra compute.
What Gemini 4 Carbon does differently
Carbon ships with a 1,048,576-token context window, so it can hold an entire codebase, a long research paper, or hours of logs in a single pass. Google says it ranks No. 1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench, and it leads Anthropic’s Claude Fable 5.2 on Terminal-Bench 4.0 with 65.2% against 61.6%.
But the numbers cut both ways. On FrontierCode 2.0, which grades real changes to open-source repositories, Fable scores 49.1% and Carbon 40.6%. The Verge flagged the same point in its coverage: on the benchmarks Google cited at the original announcement, Carbon sits behind Fable 5.2, and vendors pick the tests that flatter them. Independent tracker Artificial Analysis scores Carbon at 60 on its Intelligence Index, against 50 for Argon at the same thinking level.
Google is also previewing /swarm in Antigravity, a feature that turns one prompt into a coordinated team of agents you can track in real time. Developers can assign Carbon to the roles that need the deepest reasoning, like architectural design and auditing, while faster models handle the rest.
Carbon landed in third-party tools on day one. It is available in Cursor, GitHub Copilot for Pro+, Max, Business and Enterprise plans, Perplexity, and Devin, according to Google’s announcement.
Cybersecurity is the headline use case
Carbon did not start as a general-purpose model. Google first unveiled it on October 28 and handed it to a small group of cybersecurity defenders through its Fairwind Program, Reuters reported at the time. The two-week trial was the point. Google wanted its strongest model tested on defense before it went wide.
The results it cites are striking. Google’s Cloud Vulnerability Research team used Carbon to prove that two bugs in popular open-source software were in fact exploitable vulnerabilities, comparable in severity to the recent PHP2Shell exploit. In simulated cyberattack testing, Google says Carbon’s safeguards blocked 99% of attack trajectories from an adaptive red-teaming agent, and the per-attempt breach rate fell to 0.8%, down from 3.1% on Argon. On the Gray Swan prompt-injection benchmark, Carbon showed a 0.0% attack success rate, down from 0.7% for Argon.
DeepMind gave the model a public test of a different kind. Agents running on Carbon wrote the machine-checked proofs behind a formal verification of libsignal, the reference implementation of the Signal messaging protocol. Google says the work covered important security properties of the encryption software and took four weeks.
How much Gemini 4 Carbon costs, and the catch
The headline price is cheap by frontier-model standards. The Verge put the launch rate at about a fifth of OpenAI’s GPT-6 Astra and a quarter of Anthropic’s Fable 5.2. But cheaper tokens don’t guarantee cheaper tasks. Carbon burns more tokens per job than Argon, so a task that costs less per token can still cost more in total.
That trade-off is why Google keeps steering most users to Argon. Carbon is not on the free tier, and The Verge notes it is unclear whether lower tiers or free users will ever get it. The Gemini app dropped the older 3.1 Pro model with this launch, so Ultra subscribers now get Carbon while AI Pro subscribers keep Argon.
There is one exception to the paywall. Eligible scientists and researchers can apply for six months of free access to Google’s advanced models through Antigravity’s highest usage tier, subject to eligibility and capacity.
Why this matters
Google has now shipped two Gemini 4 frontier models in about six weeks. Argon arrived September 30, Carbon followed on October 28, and general availability came thirteen days later. The pace tells its own story. The race is no longer just about who has the smartest model. It is about who can put that model in developers’ hands fastest, cheapest, and inside the tools they already use.
For now, Carbon is a specialist tool with a general-availability label.
The real test starts this week, when developers run it on their own codebases and find out whether the extra reasoning is worth the extra tokens. Google’s own guidance gives the game away: if Argon can do the job, use Argon. Carbon earns its keep only on the jobs Argon can’t finish.
Frequently asked questions
What is Gemini 4 Carbon?
Gemini 4 Carbon is Google’s most powerful AI model, built for complex multi-step coding, cybersecurity work, and scientific research. It became generally available to paying developers and businesses on November 10, 2026.
How much does Gemini 4 Carbon cost?
Through November 30, 2026, it costs $2 per million input tokens and $10 per million output tokens. From December 1, those rates double to $4 and $20.
How is Gemini 4 Carbon different from Gemini 4 Argon?
Argon is Google’s everyday model for most coding and knowledge work. Carbon uses more thinking tokens per task and is aimed at the hardest jobs, like multi-package migrations, root-cause debugging, and open-ended research.
Who can use Gemini 4 Carbon?
Paying developers can call it through the Gemini API in Google AI Studio, and businesses can use it through Gemini for business. Google AI Ultra subscribers get it in the Gemini app, AI Mode in Search, and Antigravity. It is not on the free tier.
Sources: Reuters, The Verge, 9to5Google, Business Today, Google DeepMind blog