Gemini 4 Carbon open to everyone, with 2x pricing from December

Google has opened Gemini 4 Carbon, a model aimed at high-difficulty tasks such as long-running agent workflows, complex systems engineering and cybersecurity, with a 1M context window and up to 65,536 output tokens. It tops the CursorBench 4.0 leaderboard at 60.4% in High mode, and API pricing matches Gemini 4 Argon until November 30.

Google has opened Gemini 4 Carbon to developers, enterprises, researchers and consumers, saying it ranks first on CursorBench 4.0, WANDR and AutomationBench while completing tasks at more than 2x lower cost than the next-best external model. Access covers AI Studio, Gemini for Business, Antigravity and the AI Ultra subscription. Artificial Analysis still puts Claude Fable 5.2 ahead on its Intelligence Index, 65 to 60.

Google has opened its strongest AI model so far, Gemini 4 Carbon, to all paying developers through the Gemini API. Previously restricted to a small group of cybersecurity defenders via the Fairwind Program, the model's introductory pricing remains available until November 30.

According to ETtech, Google has expanded access to its Gemini 4 Carbon AI model for developers, enterprises, scientists, and consumers. The model brings advanced AI capabilities to complex, multi-step tasks such as software development, financial research, and cybersecurity.

ZhiDongXi reports that Google has opened its flagship Gemini 4 Carbon to developers, enterprises, researchers and consumers, callable via Google AI Studio and Gemini for business, with Google AI Ultra subscribers getting it in Antigravity, the Gemini app and AI Mode in Search. The report says it ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench.

Google has officially released Gemini 4 Carbon, its most capable reasoning model designed for complex software engineering, multi-step agent tasks, scientific research, and cybersecurity. The model is now available to developers and enterprise customers via the Gemini API, Google AI Studio, and subscription tiers.

Simon Willison released llm-gemini 0.36, which brings Gemini 4 Carbon to everyone after a two-week run limited to a subset of cyber defenders. The model is priced at $2 per million input tokens and $10 per million output tokens at launch, doubling to $4 and $20 on December 1st.

Google has made Gemini 4 Carbon generally available after a two-week testing phase, according to The Decoder. The model tops several benchmarks, beating Anthropic's new Fable 5.2 in some cases at less than half the per-task cost, and Google is previewing a /swarm feature that lets coordinated agent teams split up complex tasks in real time.

Google began rolling out Gemini 4 Carbon to paying developers and enterprises on November 10, describing it as its strongest model for multi-step coding, debugging and open-ended research, and one that spends more tokens than last month's Gemini 4 Argon on exploration and verification. Access runs through the Gemini API, AI Studio and Gemini for business, with Google AI Ultra subscribers getting it in the Gemini app, AI Mode in Search and Antigravity. Google also previewed multi-agent coordination in Antigravity and six months of free research access for 10,000 scientists.

According to cnBeta, Google has made Gemini 4 Carbon, its latest model, broadly available to the public and introduced an agent “swarm” feature. Google calls Carbon its strongest model yet, saying it can compete with top systems from Anthropic and OpenAI; two weeks ago it was limited to a small group of cybersecurity defenders.

Google has opened Gemini 4 Carbon to paying developers and businesses and is rolling it out to AI Ultra subscribers, about two weeks after a limited release to cyber defenders, The Next Web reports. It costs the same per token as Argon at launch but uses more tokens per task.

Google has opened Gemini 4 Carbon to paid API and AI Ultra subscribers, two weeks after giving the frontier model to cyber defenders, according to Google DeepMind SVP Koray Kavukcuoglu. Google also previewed a swarm feature in Antigravity.

Google has officially announced that Gemini 4 Carbon is now available to Gemini Ultra subscribers and through its APIs, according to TestingCatalog. The account added that it hopes the model will reach the Pro tier as well.

Google DeepMind has made Gemini 4 Carbon available to developers, enterprises, scientists, and consumers, pitching it at cybersecurity work and long-horizon agentic workflows across knowledge work, software engineering, and scientific reasoning. 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 it is previewing /swarm in Antigravity.

Google is rolling out Gemini 4 Carbon to the general public after two weeks of limited access for select cybersecurity defenders, The Verge reported. Google calls it its strongest model yet, aimed at complex multi-step coding and open-ended scientific research, and says the earlier Argon model remains the better pick for most software engineering and knowledge work. It is also previewing a "swarm" feature on its Antigravity agent platform that turns one prompt into a coordinated team of agents.

Google DeepMind and Artificial Analysis released the Multimodal Knowledge Agent (MKA) benchmark, a 200-task eval measuring how agents reason over visually complex files. The tasks were built with domain experts across 10 verticals and 83 occupations. Google says Gemini 4 Carbon delivers top quality on it at a fraction of the price of comparable models.

Google is opening Gemini 4 Carbon to developers, enterprises, scientists and consumers, calling it its strongest model for cybersecurity capabilities and long-horizon agentic workflows. Google says Carbon ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench at less than half the cost per task of the next best external model. Gemini 4 Argon remains the recommended choice for most software engineering and knowledge work.

Gemini 4 Carbon is rolling out on Google Cloud Platform and is already available in Agent Studio, according to TestingCatalog. The model is described as Google's most capable for deep work, with frontier performance on long-horizon agentic coding, enterprise knowledge work and scientific reasoning, plus a 1M output limit.