# Gemini 4 Carbon is broadly available: what it is for, how to get it, and how to compare cost with Argon - Category: 模型动态 - Published: 2026-11-11T19:44:29.000Z - Source date: 2026-11-10 - HTML: https://www.88z.ai/en/news/news-gemini - Markdown: https://www.88z.ai/en/news/news-gemini.md - Other language: https://www.88z.ai/news/news-gemini.md On 10 November 2026, Google said Gemini 4 Carbon is opening more widely to paying developers, enterprises, scientists, and consumers. Officially, Carbon is for longer-horizon, more demanding tasks. It shares Argon’s introductory unit price, but Google also says it uses more tokens than Argon on difficult work. ## At a glance - Carbon is opening more widely to paying developers, enterprises, and Ultra consumer surfaces; it was earlier offered to some defenders through Fairwind. - Google still points most software and knowledge work to Argon. Carbon is for harder, longer-horizon tasks and uses more tokens. - The announced introductory rate is $2 per million input tokens and $10 per million output tokens, matching Argon, and is not a relay or 88z.ai retail price. - Official benchmark rankings are vendor-disclosed. Compare completion quality and the full task bill against a fixed acceptance test. ## What opened today, and what changed two weeks ago On 10 November 2026, Koray Kavukcuoglu, Google DeepMind SVP and Google’s Chief AI Architect, said Gemini 4 Carbon is now available to developers, enterprises, scientists, and consumers. Officially, it is Google’s strongest model for cybersecurity capabilities and for long-horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning. This is not Carbon’s first appearance. The post says Carbon was released two weeks earlier to a set of cybersecurity defenders through the Fairwind Program. Gemini 4 Argon, launched before that, remains the official recommendation for the majority of software engineering and knowledge work. Carbon is described as the model for tasks that test the limits of a model’s intelligence, such as complex multi-step coding and debugging, and open-ended scientific research and discovery. ## How Google describes the work it is for For knowledge work, Google says Carbon can combine visual understanding, reasoning, and exhaustive retrieval in legal, finance, and enterprise operations, synthesizing grounded insights from large sets of dense documents and data, and following strict schemas to generate spreadsheets, presentations, and reports. The company also says existing benchmarks do not adequately capture how agents reason over dense files with charts, drawings, and tables, and that it partnered with Artificial Analysis to launch the Multimodal Knowledge Agent (MKA) benchmark. For engineering, the announcement highlights multi-package migrations, cross-subsystem root-cause investigations, and complex architectural redesigns. It says Carbon can run build-and-verify loops in web app development, recover from errors, and check both its code and the rendered UI. To apply this to larger projects, Google is previewing /swarm in Google Antigravity, turning a single prompt into a coordinated team of agents you can track in real time. The official suggestion is to assign Carbon to roles that need the deepest reasoning, such as architectural design, coordination, and auditing, and to pair it with faster, lighter models in isolated workspaces, with adversarial verifiers testing every change. ## Who can use it, and what official price was listed Access paths should be checked separately. The post says Carbon is available via the API in Google AI Studio and Gemini for business for all paying developers and enterprises. It is also rolling out to all Google AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search. The announced introductory price matches Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens. Those are Google’s announcement prices, not the final retail rates of 88z.ai or another relay. Just as important is Google’s own caveat: Carbon uses more tokens than Argon to explore thoroughly, reason deeply, and run extended verification loops. Equal unit prices therefore do not mean equal bills for a complete task. ## How to read the official scores without stopping at first place Google reports first-place results on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench, and says cost per task is more than 2 times lower than the next best external model. On MKA, it ranks second and says average cost per task is more than 6 times lower than the top-performing model. These figures come from the company’s announcement, and some of the benchmarks were launched with Google’s involvement, so they should be read as vendor-disclosed results, not independent measurements by this site. A more useful approach for buyers is to look at completion quality and the whole-task bill together. If Argon already finishes a job reliably, switching to Carbon may simply spend more tokens. If current models often fail mid-way on multi-step debugging or long-document synthesis, it is worth a small, like-for-like comparison with a fixed acceptance test. ## 88Z perspective: write the acceptance test before you move to Carbon Our suggestion is to write down the real tasks you will accept, for example a migration that must pass existing tests, or an analysis that must cite checkable sources, then run a small batch on both Argon and Carbon. Record completion, the amount of human correction, elapsed time, and total cost. That compares the full job, not a single leaderboard score. Also confirm which access path your account actually has. API access for paying customers, enterprise interfaces, and Ultra app surfaces are rolling out on separate tracks. A model name in a catalogue does not by itself mean your account can already call Carbon in the product you care about. For relay or 88z.ai prices, use the current price list rather than Google’s announced introductory rate. ## Sources - [Google](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available) (2026-11-10) - [Google · Gemini 4 Argon](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon) (2026-10-13) This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.