Google on November 10 began rolling out Gemini 4 Carbon, the company's self-described "strongest model" yet, to Google AI Ultra subscribers across the Gemini app, AI Mode in Search, and AntiGravity, its agentic development platform1,2,3. The model is also available to developers and enterprises through the Gemini API and Google AI Studio4,6. Carbon had been available only to a select group of cybersecurity defenders since late October through Google's Fairwind Program.
Google said Carbon "pushes the frontier even further" than Argon, the first Gemini 4 model, which launched only a month prior. The company said it considers Carbon capable of competing with the leading systems from Anthropic and OpenAI. Google cited independent benchmark results: Carbon ranks first on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench at what Google described as more than two times lower cost per task compared to the next best external model.
Where Carbon and Argon split
Google drew a clear line between the two Gemini 4 variants. Argon, available to AI Pro subscribers and above, will remain what Google called "the ideal choice for the majority of software engineering and knowledge work". Carbon targets what Google described as tasks that test the limits of a model's intelligence: complex, multi-step coding and debugging, and open-ended scientific research and discovery — using more tokens than Argon to explore thoroughly, reason deeply, and run extended verification loops. In the Gemini app's model picker, Argon is labeled for "Code & analysis" while Carbon is labeled for "Most complex work". Google removed Gemini 3.1 Pro from the Gemini app with this launch.
For legal, finance, and enterprise work, Google said Carbon brings "higher-precision knowledge work, artifact generation, and data analysis". The model is also available in Gemini for business.
Swarm preview and early results
Alongside the broader rollout, Google previewed a "swarm" feature on AntiGravity that turns a single prompt "into a coordinated team of agents you can track in real time"5. Developers can assign Carbon to roles demanding deep reasoning, such as architectural design, coordination, and auditing, while pairing it with faster, lighter models across parallel isolated workspaces and adversarial verifiers that test every change.
Google said Carbon had helped its first users uncover zero-day vulnerabilities in a short span of time and had made advances in protein structure prediction. The company is giving 10,000 scientists free access to Carbon for six months to support their research and said it has already onboarded thousands.
ANALYSIS The two-tier Gemini 4 lineup — Argon for breadth, Carbon for depth — pairs a cost-efficient workhorse with a premium reasoning model, but the swarm feature adds a distinct wrinkle: it positions AntiGravity as an orchestration layer where Carbon and lighter models collaborate rather than compete for the same workload. The benchmark-cost claim, if it holds under independent scrutiny, would give Carbon a concrete economic argument beyond raw accuracy.