Google has made Gemini 4 Carbon, its strongest AI model yet, available to developers, enterprises, scientists and consumers, two weeks after limiting the model to a small group of cybersecurity defenders. The wider release, announced on Tuesday, also brings a preview of /swarm in Google Antigravity, a feature that turns a single prompt into a coordinated team of AI agents.
From cyber defenders to everyone
Carbon was unveiled on October 28 and first offered through Google's Fairwind Program to a subset of cybersecurity defenders. From today, all paying developers and enterprises can use it through the Gemini API in AI Studio and Gemini for business, at the same introductory price as Gemini 4 Argon — $2 per million input tokens and $10 per million output tokens. Google AI Ultra subscribers also get Carbon in the Gemini app, AI Mode in Search and Antigravity. Google says Argon remains the better fit for most software engineering and knowledge work, while Carbon is built for tasks that push a model's limits: complex, multi-step coding and debugging, and open-ended scientific research. Carbon uses more tokens than Argon to explore thoroughly, reason deeply and run extended verification loops.
Benchmarks and the /swarm preview
Google claims Carbon ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench, at more than twice lower cost per task than the next best external model. On a new Multimodal Knowledge Agents benchmark built with Artificial Analysis, it ranks second while costing more than six times less per task than the top model.
The headline addition for developers is /swarm, now in preview in Antigravity. A single prompt becomes a team of agents that can be tracked in real time: Carbon can be assigned the deepest-reasoning roles, such as architectural design, coordination and auditing, while faster, lighter models work in parallel in isolated workspaces, with adversarial verifiers testing every change.
Cyber, science and safety
Carbon's cybersecurity pitch is its strongest claim. Through the Fairwind Program it leads benchmarks for finding, exploiting and patching vulnerabilities, and in Google's Project Fortify system it hit 100% recall on scanning and 95% precision on triage in internal testing. Google says its Cloud Vulnerability Research team used Carbon to prove two bugs in popular open-source software were exploitable, getting them reclassified as critical, and Wiz's Scan for Good programme found zero-day flaws at two major SaaS providers in under two hours.
In science, Google points to drug-toxicity research at Axiom Bio, where Carbon is now the leading model for translational science, and to researchers using it to help formally verify the code behind Signal's cryptographic protocol — work that historically takes expert teams years. Google is also expanding free access to Carbon to the 10,000 scientists in its Antigravity research programme, free for six months.
On safety, Google says Carbon blocked 99% of attack trajectories in simulated cyberattack tests, cutting the per-attempt breach rate to 0.8% from 3.1% on Argon, and recorded a 0% prompt-injection success rate on Gray Swan's benchmark. On an agentic honesty benchmark it passed 92% of tasks, up from 70% for Argon.



