
13 Oct 2026
Google releases Gemini 4 Argon to everyone who pays, two weeks after holding it back for cyber defenders
Google on Tuesday, Oct. 13, 2026, made Gemini 4 Argon, its new frontier model, globally available to developers and enterprises through the Gemini API, to Google AI Ultra subscribers in its Antigravity coding platform, and to Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search. Google first announced Argon on Sept. 30 but limited it to trusted cyber defenders while it hardened safeguards. Google says Argon ranks first on the Vals Index of finance, coding, legal and tax work.
For most of this year the frontier race looked like a two-company fight, and Google was the giant standing on the sidelines. Argon is Google's answer, and the most aggressive part of it may be the price: a fifth of what rivals charge, at least until December. The other thing worth noticing is the order of operations. Google gave the model to cyber defenders first, watched it find real holes (including one guarding hundreds of thousands of patient records), and only then opened it up, wrapped in monitors that literally read the model's internal signals for weapons knowledge. That's the new playbook for frontier launches. The catch is in Google's own safety post: Argon cleared all 12 of its simulated cyberattack scenarios. The same skill that patches hospital software can break into it, which is exactly why the guardrails are the product now.
On Tuesday, Oct. 13, 2026, Google said Gemini 4 Argon is globally available. Developers can use it on the paid tier of the Gemini API in Google AI Studio and Android Studio. Enterprises can use it in Gemini for business, the API and Google Antigravity, Google’s coding platform for AI agents. Google AI Ultra subscribers get it in Antigravity and AI Studio. Google AI Pro and Ultra subscribers get it in the Gemini app and in AI Mode in Search. The Verge reports it is not clear when, or if, Argon will reach Plus subscribers or free users. Those availability lines are Google’s. The Plus and free-user caveat is The Verge’s.
Google first announced Argon on Sept. 30, 2026, but released it only to trusted cyber defenders through its Fairwind Program while it strengthened safeguards. A cyber defender, here, is someone whose job is to find and fix attacks. On Oct. 13 Google said it is expanding Fairwind to hundreds of additional partners. Fairwind members can apply for access to Argon without cyber guardrails — the limits that stop a model from helping with an attack. Those lines are Google’s.
Introductory pricing is $2 per million input tokens and $10 per million output tokens, with cached input 95% off. A token is a small piece of text the model reads or writes. Cached input is text the service already stored from an earlier request, so a later call can reuse it. The Verge reports that introductory price is around a fifth of the cost of OpenAI’s GPT-6 Astra and Anthropic’s Fable 5 models, and that prices double at the end of November when the introductory period ends. The dollar rates are Google’s. The fifth-of-rivals comparison and the November end date are The Verge’s.
Google says Argon ranks No. 1 on the Vals Index, a test of finance, coding, legal and tax work, and claims it outperforms rival models including GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5 on benchmarks across coding, cybersecurity, science, finance and law. That ranking is Google’s, as The Verge reports it. Argon has a 1 million token unified input-output limit via the Gemini API, meaning the model can read and write up to that many tokens in one run. That ceiling is Google’s.
Google’s internal examples: Argon in Antigravity raised agentic code submissions by more than 35% and cut code rollbacks by nearly a third. Agentic, here, means the software takes a chain of steps on someone’s behalf, not only answers a question. A rollback is a change that has to be undone. Google says Argon helped the Chrome team untangle a roughly 6 million-line legacy folder inside Chromium’s 40 million-line codebase. Chromium is the open-source browser project Chrome is built on. Working overnight, Google says, it found optimizations now live in Google’s open-source Agent Development Kit that cut in-memory session prep time 40% in long sessions. It also delivered a 2.85x inference speedup for Google’s open Gemma 4 E2B model on an Nvidia A100 chip using the open vLLM serving software. Those results are Google’s account of its own systems.
Google says security firm Wiz, with early access, used Argon to find and fix a public file-download bypass in healthcare software that exposed a master key and administrator access, protecting hundreds of thousands of patient records. Those lines are Google’s.
In the Gemini app, Google says Argon answers more directly “with less hedging.” The app adds a chat-to-tasks toggle so Gemini can keep working when you’re away, Antigravity’s coding interface inside the Gemini macOS app, and connections to more than 40 apps including Instacart, OpenTable, Adobe and Shopify. Those lines are Google’s.
On safety, Google says Argon solved all 12 of its end-to-end simulated cyberattack scenarios, up from five for the previous generation. Unguarded versions showed much stronger ability to troubleshoot complex lab procedures, so Google applied its next-generation safeguards for cyber and biology under its Frontier Safety Framework, the company’s rules for its most capable models. External bio jailbreak red-teaming unlocked less than 1% of tested bio sub-tasks, and violations on an adversarial test set fell nearly twofold. A jailbreak, here, is a prompt designed to make a model ignore its limits. Google says the safeguards include input screening, “activation probes” that read the model’s internal signals for cyberattack or bioweapon knowledge, and continuous monitoring that can cut an agent off mid-stream. Those lines are Google’s.
The Verge notes Argon is Google’s first major frontier release since Gemini 3 in late 2025, and it lands a day after Meta announced Muse Wonder Full. Those lines are The Verge’s.
The picture is Google’s official Gemini 4 Argon launch artwork: the Gemini logo and the words Gemini 4 Argon beside a large translucent number 4 on a pale blue background. It is the company’s launch graphic. The frame does not print a calendar date.
In plain terms, Google said on Tuesday that Gemini 4 Argon is now open to paying developers, enterprises, and Pro and Ultra subscribers, two weeks after a limited release to cyber defenders. The introductory price is $2 and $10 per million tokens. Google says it ranks first on a finance-coding-legal-tax index and that Wiz used it to close a hospital-software hole. The Verge says Plus and free users are still waiting, and that the cheap price lasts until the end of November.
RELATED
- Google announces Gemini 4 Argon, limited to trusted cyber testers for now
- Meta launches Muse Wonder, its most capable AI models yet, and says it has caught up to the frontier
- Britain says its AI testers are now evaluating Gemini 4 Argon, and tells AI labs: if you can't control it, don't release it
- OpenAI published GPT-6 Astra on its own site
- Anthropic published Claude Fable 5.1 and Mythos 5.1 — same model, different safeguards
- Anthropic launches Claude Opus 5.5
Sources
- Google — Gemini 4 Argon: Built to solve your hardest problems, 13 Oct 2026
blog.google
- Google — Gemini 4 Argon is now available for developers, 13 Oct 2026
blog.google
- Google — Gemini 4 Argon comes to the Gemini app, 13 Oct 2026
blog.google
- Google — Developing Gemini 4 Argon responsibly and safely, 13 Oct 2026
blog.google
- The Verge — Google finally releases Gemini 4 Argon, 13 Oct 2026
theverge.com