Dept. of Mysteries
Gemini 4 Carbon Arrives as Google Unveils Its Strongest Model in Wondrous Fashion
Google has made Gemini 4 Carbon, its strongest model, available to developers and enterprises at $2 per million input tokens and to Google AI Ultra subscribers.
Tuesday, 10 November 2026 · 2 min read
Google has made Gemini 4 Carbon available to developers, enterprises, scientists and consumers, declaring it the company's strongest model. The release matters because Carbon is built for the most arduous labours a machine can attempt: long, multi-step coding and debugging, open-ended scientific research, and agentic workflows, meaning tasks the model carries out over many steps with a degree of independence. Two weeks earlier, Google had released Carbon to a set of cybersecurity defenders through its Fairwind Program. "Carbon pushes the frontier even further," said Koray Kavukcuoglu, senior vice president at Google DeepMind and Google's chief AI architect. Carbon's sibling, Gemini 4 Argon, remains the recommended choice for most software engineering and knowledge work, while Carbon spends more tokens, the small units of text a model reads and writes, to explore thoroughly and run extended verification loops. Google reports that 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. By decree of its makers, the model is available via the API in Google AI Studio and Gemini for business for all paying developers and enterprises, at the same introductory price as Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens. It is also rolling out to all Google AI Ultra subscribers in Google Antigravity, the Gemini app and AI Mode in Search. Alongside it, Google is previewing /swarm in Google Antigravity, a feature that turns a single prompt into a coordinated team of agents that can be tracked in real time. On security, an arcane art if ever there was one, Google says Carbon leads on key benchmarks for finding, exploiting and patching vulnerabilities when accessed through the Fairwind Program. In Project Fortify, Google's autonomous system for scanning, triage and remediation, Carbon hit 100% recall on scanning and 95% precision on triage on an internal benchmark, recall being the share of real flaws a system catches and precision the share of its verdicts that prove correct. As part of Wiz's Scan for Good program, the model uncovered 0-day vulnerabilities, flaws unknown to the software's makers, in under two hours while scanning two major SaaS (software-as-a-service) providers, firms that deliver software over the internet. Behold, too, its scientific feats. Google researchers used Carbon within a new formal verification framework called AlphaProof Code to help prove key security properties of the code underlying the cryptographic protocol of Signal in weeks, a task that historically takes expert teams years. Drug toxicity research company Axiom Bio, meanwhile, is moving its production scientific analyses and agents to both Carbon and Argon, a fresh entry in the annals of machine-aided discovery.
Source: blog.google