Google opens Gemini 4 Carbon to developers at $2/$10

Google on Tuesday, November 10, opened Gemini 4 Carbon to paying developers, enterprises, scientists and Google AI Ultra subscribers. Carbon is the company's most powerful AI model so far. The wider release came about two weeks after Google limited access to cybersecurity defenders in its Fairwind Program, according to Google's announcement. Alphabet's ↘ 0.60% Google is pricing the model to win developers. It costs $2 per million input tokens and $10 per million output tokens through November 30, then rises to $4 and $20.blog.google

Access and pricing

Carbon is available through the Gemini API in Google AI Studio and Gemini for business for all paying developers and enterprises. It is also rolling out to Google AI Ultra subscribers in the Gemini app, AI Mode in Search and the Antigravity coding platform, Google said. The Verge reported that it is not clear when, or if, lower-tier subscribers or free users will get the model.theverge.com+1

The introductory rate is about a fifth of the cost of OpenAI's GPT-6 Astra and a quarter of Anthropic's Fable 5.2, which launched November 9, according to The Verge. The outlet noted that lower per-token prices do not always mean lower costs, because models can use different numbers of tokens for the same task. VentureBeat reported that Carbon's price after the promotion will match Anthropic's Claude Opus 5.5 and stay at half of Fable 5.2's standard rates.venturebeat.com+1

Argon stays the default

Google said Gemini 4 Argon, which it released last month, "remains the ideal choice for the majority of software engineering and knowledge work." It described Carbon as "built for tasks that test the limits of a model's intelligence," such as complex multi-step coding and open-ended scientific research, and said Carbon uses more tokens to reason and verify its work, according to the company's post. In the Gemini app's model picker, Argon is labeled for "Code & analysis" and Carbon for "Most complex work." Gemini 3.1 Pro was removed from the app with this launch, 9to5Google reported.9to5google.com+1

Google also previewed /swarm in Antigravity. The feature turns a single prompt into "a coordinated team of agents you can track in real time." Developers can assign Carbon to roles such as architecture and auditing and pair it with lighter models, Google said.blog.google

Benchmarks and safety claims

Google says Carbon ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench, at more than 2X lower cost per task than the next-best external model. Google also expanded its free six-month research program to give scientists access to Carbon. On safety, Google said its safeguards blocked 99% of attack attempts in simulated cyberattack tests, per the announcement.blog.google

The Verge cautioned that companies "can and do cherry-pick benchmark results." The outlet said Google's earlier benchmarks put Carbon behind only Fable 5.2 on most measures, and it called the release a return for Google to the front of AI development after months of competition dominated by Anthropic and OpenAI.theverge.com