Gemini 4 Carbon opens paid API and Ultra access, introductory pricing ends November 30

As of November 10, 2026, Gemini 4 Carbon is available through multiple channels beyond the initial [Fairwind Program](https://blog.google/innovation‑and‑ai/models‑and‑research/gemini‑models/gemini‑4‑carbon‑now‑available/) for cybersecurity defenders [1]. Paid API access is live in Google AI Studio and Gemini for business for all paying developers and enterprises. Google AI Ultra subscribers can use Carbon in Google Antigravity, the Gemini app, and AI Mode in Search. Scientists and researchers can [apply for access](https://blog.google/innovation‑and‑ai/models‑and‑research/gemini‑models/gemini‑4‑carbon‑now‑available/) to both Argon and Carbon via Google Antigravity's highest usage tier free for six months, subject to eligibility and capacity [1]. Everyday users with Ultra subscriptions can try Carbon in the Gemini app and AI Mode in Search.

During the introductory period, Carbon carries the same per‑token pricing as Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens [1]. This rate applies to "the Carbon model intended for public access only" and expires on November 30, 2026. After that date, the price doubles to $4 per million input tokens and $20 per million output tokens [1]. Google states Carbon "uses more tokens than Argon to explore thoroughly, reason deeply, and run extended verification loops" [1], meaning actual workload costs may be higher even at identical per‑token rates during the introductory window.

Google positions Argon as "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" [1]. According to the announcement, Carbon targets complex multi‑step coding and debugging, open‑ended scientific research, and cybersecurity workflows requiring deeper reasoning. The [OpenTools model catalogue](https://opentools.ai/llm) provides comparison data for alternative models [2]. For most development tasks where extended verification loops aren't necessary, Argon's lower token consumption may offer better cost efficiency. The earlier [Carbon access article](https://opentools.ai/news/gemini‑4‑carbon‑trusted‑defenders‑access‑introductory‑later‑token‑pricing) documented the restricted cybersecurity rollout that preceded this general availability expansion [3].

Google claims Carbon shows "frontier performance in cybersecurity capabilities and long‑horizon agentic workflows" [1] and ranks highly on benchmarks including CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench. The model is described as having stronger alignment—passing 92% of tasks on Google's 375‑task agentic honesty benchmark compared to 70% for Argon—and improved safety mitigations for cybersecurity applications [1]. However, the announcement notes Carbon "uses more tokens than Argon" and that "prompt injection robustness remains an area for continued improvement" despite achieving a 0% success rate on the Gray Swan IPI benchmark [1]. Access remains selective for cybersecurity defenders through Fairwind and for researchers through the scientific program.