Google rolls out Gemini 4 Carbon to paid API and AI Ultra users two weeks after releasing it to cyber defenders, and previews a swarm feature in Antigravity
First reported by Blog.google ·
The top-tier AI model is now available to all paying developers and enterprise customers.
Google has made its Gemini 4 Carbon AI model available to paid API users and AI Ultra subscribers, two weeks after initially releasing it to cybersecurity professionals. This advanced model, described as the company's strongest, excels in frontier cybersecurity capabilities, long-horizon agentic workflows for complex knowledge work, software engineering, and scientific reasoning. Carbon is positioned for tasks that push the limits of AI intelligence, such as intricate multi-step coding and debugging, and open-ended scientific research, utilizing more tokens than its predecessor, Gemini 4 Argon, for deeper exploration and verification. It also offers competitive performance and cost-effectiveness on benchmarks like CursorBench 4.0 and Zapier's AutomationBench. Google also previewed a new "/swarm" feature in its Antigravity platform, which allows a single prompt to orchestrate a team of agents, including Gemini 4 Carbon, for complex tasks like architectural design and auditing. The model is accessible via the API in Google AI Studio and Gemini for business at the same introductory pricing as Gemini 4 Argon, and is rolling out to Google AI Ultra subscribers across various Gemini applications and AI Mode in Search.
Gemini 4 Carbon's broad release signifies Google's strategy to democratize access to its most powerful AI capabilities, moving frontier models from specialized programs to wider developer and enterprise adoption. This includes significant advancements in multimodal understanding and agentic workflows, aiming to tackle complex tasks across various industries from finance and legal to 3D design and cybersecurity. The introduction of the /swarm feature, allowing coordinated agent teams, hints at a future where AI operates more autonomously and collaboratively on sophisticated projects, potentially accelerating development cycles and creative output.
The model's performance claims on benchmarks like CursorBench and Zapier's AutomationBench, coupled with competitive pricing, suggest a move towards balancing cutting-edge AI power with practical production economics for businesses. This positions Gemini 4 Carbon as a direct competitor for demanding AI workloads, potentially impacting the competitive landscape for AI-powered development tools and enterprise solutions. Developers and enterprises should evaluate Carbon for complex knowledge synthesis, sophisticated coding, and intricate design tasks where its advanced reasoning and multimodal capabilities can offer distinct advantages.
AI-written summary. May contain errors.