News · Available now
Google releases Gemini 4 Argon: what changes for Gemini app subscribers and teams building on it
The short version
Gemini 4 Argon is out now for Google AI Pro and Ultra subscribers and on the paid Gemini API, at introductory prices until November 30, 2026.
Google released Gemini 4 Argon, its new frontier model, on October 13, 2026, and you can use it now. If you already pay for Google AI Pro or Ultra, Argon is in the Gemini app and AI Mode in Search. If your team builds on Google's API, it's on the paid tier at an introductory price that doubles after November 30.
Who gets it, and where
- Google AI Pro and Ultra subscribers: the Gemini app and AI Mode in Search.
- Google AI Ultra subscribers: also Google Antigravity, Google's agent-first development platform, and AI Studio.
- Developers: the paid tier of the Gemini API in Google AI Studio and Android Studio.
- Enterprises: Gemini for business, the API and Google Antigravity.
Google says Argon is built for long, multi-step work in coding, finance, legal, cybersecurity and science. It also says Argon ranks #1 on the Vals Index, which measures performance on finance, coding, legal and tax work. Those are Google's own benchmark claims.
What changes in the Gemini app
According to the Gemini app post, the model isn't the only change. A toggle in the main chat window switches from chat to tasks, and Google says Gemini can keep working on tasks while you're away from your computer. Gemini now connects to more than 40 apps, including Instacart, OpenTable, Adobe and Shopify. The Gemini macOS app now includes Antigravity's agentic coding interface.
Google also says Argon writes differently: "Argon answers questions more directly with less hedging, doesn't repeat itself more than necessary, and skips clichés throughout the responses." On longer tasks, it sends progress updates more often.
What it costs on the API
The launch post lists an introductory price of $2 per million input tokens and $10 per million output tokens. Tokens are the small chunks of text a model reads and writes. Cached input tokens are 95% off. After the introductory period ends on November 30, 2026, the price rises to $4 per million input tokens and $20 per million output tokens.
Through the Gemini API, Argon has a combined input and output limit of 1 million tokens. Google points developers to its documentation for a new "continuation protocol" and setup details. The developer post links to the Gemini API Developer Guide.
Safety notes that matter if you use agents
Google's launch post and safety post say Argon was trained to stay within system boundaries, report negative findings honestly and ask for human confirmation before irreversible steps. Per Google, attempts to take shortcuts on coding benchmarks dropped to 2%. On Gray Swan's Indirect Prompt Injection benchmark, attack success fell to 0.7% across 15 adaptive attempts. In prompt injection, hidden instructions in webpages, emails or code try to take over an agent's task.
What it means for your business
If you're on Google AI Pro, Argon is already part of your plan, so test it on real work this week. Google's own example is a good place to start. In the Gemini app, Argon takes a messy folder of project notes, uses a custom skill to turn it into a client proposal deck, drafts and emails the deck, and sets up a scheduled task for future client requests. A small agency or consultancy could run that on one live proposal. Stop before the email goes out and read the deck yourself. A 0.7% attack success rate is low, but it isn't zero. An agent that emails your clients is exactly where one mistake costs you.
Check which plan your team actually has. Pro gets Argon in the app and Search. Antigravity is listed for Ultra subscribers and enterprises.
If you build on the API, budget at the post-November price, not the introductory one. Here's an illustration using Google's rates: a workload of 50 million input tokens and 10 million output tokens a month costs $200 now and $400 from December. If your prompts reuse the same long context, such as a policy manual or product catalog, caching that input is where the 95% discount helps.
Things to watch: Google says this is the first of the Gemini 4 family, and it hasn't given dates for the other models.
Sources
- Gemini 4 Argon is now available for everyone · Googlehttps://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/
- Gemini 4 Argon is now available in the Gemini app · Googlehttps://blog.google/innovation-and-ai/products/gemini-app/gemini-4-argon-availability/
- Gemini 4 Argon is now available for developers · Googlehttps://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-4-argon/
- Developing Gemini 4 Argon responsibly and safely · Googlehttps://blog.google/innovation-and-ai/technology/safety-security/gemini-4-model-safety-security/
Drafted by Elle Evate, our Head of AI Content (an AI system). Reviewed and approved by a Persistent Momentum editor. How we work: /editorial-policy. Spot an error? /corrections.