Google expanded access to Gemini 4 Argon on October 13, 2026, bringing the model to paid Gemini users and developers through several products. But app subscriptions and developer access have different terms.
The wider release follows an initial restricted rollout announced on September 30. Google says Argon is designed for complex coding and knowledge work, including analysing and combining information from documents.
Which plan covers access?
Gemini app and AI Mode in Search: Google announced access for Pro and Ultra subscribers. Its current plans page lists Argon as available to Ultra and rolling out to Pro subscribers in selected countries—not an assurance that every account has it yet.
Google AI Studio: The browser-based development interface includes Argon through an Ultra subscription, with usage limits. Its Free and Pro subscription options do not include Argon. That restriction is specific to AI Studio, not Pro access in the Gemini app.
Google Antigravity: Google’s launch announcement includes Argon access for Ultra subscribers using this coding tool.
Direct Gemini API: Usage from external applications is billed separately from AI Studio subscription allowances. Eligible Google Cloud credits may offset charges under their own rules; an app subscription is not unlimited API access.
Before upgrading specifically for Argon, check the product and account you intend to use. The documentation checked does not establish a Pro rollout completion date. Teams moving from older coding subscriptions can also consult our Code Assist renewal and Antigravity migration guide.
API rates double on December 1
Google’s Standard API rate card lists the following prices in US dollars per million tokens. Tokens are the units used to count model input and generated content; output billing includes the model’s thinking, not just the answer you see.
Standard API usage | Through November 30, 2026 | From December 1, 2026 |
|---|---|---|
Uncached input | $2 | $4 |
Output, including thinking | $10 | $20 |
Calculated example: 100,000 uncached input tokens plus 10,000 billed output tokens, including thinking, cost $0.20 + $0.10 = $0.30 at introductory rates. The same token usage costs $0.40 + $0.20 = $0.60 from December 1. This is arithmetic using Google’s published rates, not a measured task. It excludes tools, retries, taxes and other fees, and assumes no cache discount.
For an application expected to run beyond November, budget against the December rates as well as the introductory price.
A million tokens is a shared budget
Argon’s API has a 1,048,576-token shared window for input, thinking and output. More source material or reasoning leaves less space for the answer; it is not a promise of a million-token response. The model specification lists a default synchronous output limit of 65,536 tokens, with longer generation requiring the long-decode feature.
Google’s model card also notes that Argon can invent information, respond slowly or time out. Its knowledge cutoff is March 2026. For document-heavy work, check important conclusions against the supplied sources rather than treating the larger capacity as proof of accuracy.
Methodology: AI-assisted reporting and analysis based on Google’s published announcements and documentation, checked on October 13, 2026. The cost example is a calculation; no hands-on model testing or independent benchmark reproduction was performed.