Google's Gemini 4 Carbon goes on sale to paying API users at $2 and $10
Google DeepMind announced on Nov 10, 2026 that Gemini 4 Carbon is available to developers, enterprises, scientists and consumers. The post was written by Koray Kavukcuoglu, Google's Chief AI Architect. Carbon had gone two weeks earlier to a set of cybersecurity defenders through the Fairwind Program.
API access runs through Google AI Studio and Gemini for business for all paying developers and enterprises, at an introductory $2 per million input tokens and $10 per million output tokens, the same as Argon. Google AI Ultra subscribers are getting it in Google Antigravity, the Gemini app and AI Mode in Search. A /swarm feature, previewed in Antigravity, turns one prompt into a coordinated team of agents.
Google positions Carbon for harder multi-step coding and scientific work and says it uses more tokens. Argon, launched recently, remains its pick for most software engineering and knowledge work. Google says Carbon ranks #1 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, and #2 on the new Multimodal Knowledge Agent benchmark at more than 6X lower average cost than the top model. In science, it retrieved on average 40 times more homologs across 7 low-homolog protein families from the CASP17 blind test set. Google researchers also used it with AlphaProof Code to generate millions of lines of Lean and prove key security properties of Signal's cryptographic protocol code in weeks.
Why it matters: the cost per task is the figure to watch. If frontier-level work on hard coding, security and research problems is available at a price flat with the previous model, more people can afford it. Google also opened applications for 10,000 scientists to get free Argon access for six months.
Caveats: nearly all performance claims are Google's own. The Project Fortify security results come from its internal benchmark. The protein work is still being benchmarked more broadly, and the cyber benchmark leads apply to Fairwind Program access. Independent testing is the next thing to look for.
Source: Google DeepMind
Many Minded