{"slug": "googles-strongest-ai-model-is-out-for-paying-users-a-day-after-claude-fable-5-2", "title": "Google’s strongest AI model is out for paying users, a day after Claude Fable 5.2 took its crown", "summary": "Google released Gemini 4 Carbon to paying developers, businesses and Google AI Ultra subscribers on Tuesday, two weeks after limiting its strongest model to cyber defenders, and one day after Anthropic's Claude Fable 5.2 displaced it atop Artificial Analysis's Intelligence Index. Claude Fable 5.2 scores 65 on the index at $26.54 per task, versus 60 for Gemini 4 Carbon at $7.76 per task, and Google paired the release with /swarm, an Antigravity preview tool that turns one prompt into a team of coordinating AI agents with adversarial verifiers. Carbon carries API pricing of $2 per million input tokens and $10 per million output tokens until November 30, rising to $4 and $20, a 1 million token context window and a March 2026 knowledge cutoff; free API users and AI Pro subscribers do not get access.", "body_md": "Google’s artwork for the general release of Gemini 4 Carbon. Image: Google\n\nGoogle has [made Gemini 4 Carbon available](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/) to paying developers, businesses and its top-tier subscribers, two weeks after [giving its strongest model only to cyber defenders](https://madrobot.blog/2026/10/28/google-gemini-4-carbon-tops-leaderboard-cyber-defenders/). It comes with /swarm, a preview tool that turns one prompt into a team of AI agents, and lands a day after Anthropic’s Claude Fable 5.2 knocked it off the top of the leading independent ranking.\n\n## Who can use Gemini 4 Carbon, and where\n\nCarbon went live on Tuesday through several routes, Google says:\n\n- **Developers:** the paid tier of the Gemini API, in Google AI Studio and Android Studio. Free API users don’t get it.\n- **Businesses:** Gemini for business, through the API or its model menu, plus Google Antigravity and CodeMender, Google’s code security agent.\n- **Subscribers:** Google AI Ultra only, in the Gemini app, AI Mode in Search and Antigravity. AI Pro subscribers keep[Gemini 4 Argon, which reached them on October 13](https://madrobot.blog/2026/10/13/google-gemini-4-argon-available-ai-pro-ultra-api-robotics/) .[9to5Google reports](https://9to5google.com/2026/11/10/gemini-app-4-carbon/) that the Gemini app’s model menu now labels Carbon “Most complex work” and has dropped Gemini 3.1 Pro.\n- **Scientists:** Google’s free six-month research programme, which opened to 10,000 scientists with Argon, now includes Carbon too.\n- **Cyber defenders:** the specialist “Carbon for cyber” stays in Google’s[Fairwind Program](https://deepmind.google/fairwind-program/) .\n\nGoogle’s [AI plans page](https://one.google.com/about/google-ai-plans/) says Carbon reaches Ultra subscribers “in select countries” but doesn’t list them, and Google hasn’t said whether the EU is included. Ultra itself is sold in more than 150 countries and starts at $99.99 a month in the US.\n\n## What /swarm does\n\nThe headline feature lives in Antigravity, Google’s agent coding app. Typing /swarm hands a prompt to a coordinator agent, which [spins up parallel workers](https://antigravity.google/blog/gemini-4-carbon-in-google-antigravity/) in separate copies of the code so their edits never collide, plus “adversarial verifiers” that run tests and challenge every change. The agents talk on a shared message board and, Google says, sit idle at no token cost until they’re needed. A side panel shows the debate live and lets you jump in to steer any agent.\n\nEach agent can run a different model. Google’s pitch is Carbon for the hardest roles (planning, coordinating, auditing) and Argon or lighter models for routine edits. It warns that swarms “consume quota much faster than a standard chat”. Its demos include a 3D game in which a pelican rides a bike around Google’s Mountain View campus, built by a swarm of Carbon agents. /swarm is rolling out in preview in the Antigravity app and command-line tool.\n\n## What Gemini 4 Carbon costs\n\nAPI prices match Argon’s: **$2 per million input tokens and $10 per million output tokens** until November 30, then $4 and $20. Carbon is built to think for longer than Argon, though, so the same job costs more. Running [Artificial Analysis’s index](https://artificialanalysis.ai/models/gemini-4-carbon), it produced 205 million tokens, almost three times the median, and cost $7.76 per task at the full price. The model card gives it a 1 million token context window and a March 2026 knowledge cutoff.\n\n## How it compares with Claude and GPT\n\nWhen Carbon was unveiled it led Artificial Analysis’s Intelligence Index by a point. It now scores 60, but [Claude Fable 5.2](https://madrobot.blog/2026/11/09/claude-fable-5-2-released-price-benchmarks/), released on Monday, has moved above it:\n\n| Model (setting) | Index score | Cost per task | \n|---|---|---|\n| Claude Fable 5.2 (max) | 65 | $26.54 | \n| Claude Fable 5.2 (high) | 62 | $9.08 | \n| Gemini 4 Carbon (high) | 60 | $7.76 | \n| Claude Opus 5.5 (max) | 59 | $10.04 | \n| GPT-6 Astra (max) | 57 | $6.24 | \n\nArtificial Analysis Intelligence Index, as listed on November 10, 2026.\n\nGoogle picks its own fights. It says Carbon ranks first on CursorBench 4.0, Perplexity’s WANDR research test and Zapier’s AutomationBench, at less than half the cost per task of the next best outside model; Anthropic’s own figures also had Carbon ahead of Fable 5.2 on AutomationBench. Google also says Carbon blocked 99% of attack attempts in simulated cyberattack tests and passed 92% of its honesty tasks, against 70% for Argon. None of Google’s figures have been checked independently.\n\nGoogle DeepMind’s Koray Kavukcuoglu, who wrote the announcement, says most people won’t need Carbon:\n\nArgon remains the ideal choice for the majority of software engineering and knowledge work, while Carbon is built for tasks that test the limits of a model’s intelligence.\n\nKoray Kavukcuoglu, Chief AI Architect, Google\n\n## Why it matters\n\nGoogle has gone from a cyber-only preview to paying customers in two weeks, but it has lost the top spot on the way. Carbon’s case now rests on price and stamina rather than raw rank, and /swarm is Google’s bet that running whole teams of agents, not one chatbot, is where the top models earn their keep.\n\n*Sources: [Google](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/), [Google Antigravity](https://antigravity.google/blog/gemini-4-carbon-in-google-antigravity/), [Gemini 4 Carbon model card](https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-4-Carbon-Model-Card.pdf), [Google AI plans](https://one.google.com/about/google-ai-plans/), [Artificial Analysis](https://artificialanalysis.ai/leaderboards/models), [9to5Google](https://9to5google.com/2026/11/10/gemini-app-4-carbon/)*", "url": "https://wpnews.pro/news/googles-strongest-ai-model-is-out-for-paying-users-a-day-after-claude-fable-5-2", "canonical_source": "https://madrobot.blog/2026/11/10/google-gemini-4-carbon-available-ai-ultra-api-swarm/", "published_at": "2026-11-10 17:01:00+00:00", "updated_at": "2026-11-10 17:21:40.128558+00:00", "lang": "en", "topics": ["large-language-models", "generative-ai", "ai-agents", "ai-products", "ai-tools"], "entities": ["Google", "Gemini 4 Carbon", "Anthropic", "Claude Fable 5.2", "Google Antigravity", "Artificial Analysis", "Gemini 4 Argon", "Google AI Ultra"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/googles-strongest-ai-model-is-out-for-paying-users-a-day-after-claude-fable-5-2", "markdown": "https://wpnews.pro/news/googles-strongest-ai-model-is-out-for-paying-users-a-day-after-claude-fable-5-2.md", "text": "https://wpnews.pro/news/googles-strongest-ai-model-is-out-for-paying-users-a-day-after-claude-fable-5-2.txt", "jsonld": "https://wpnews.pro/news/googles-strongest-ai-model-is-out-for-paying-users-a-day-after-claude-fable-5-2.jsonld"}}