Google made Gemini 4 Argon generally available through its paid API on October 13, two weeks after introducing it to a restricted group of cybersecurity testers. The new gemini-4-argon model ID gives developers a documented way to integrate and evaluate it. Access through Google's consumer app follows different subscription rules, while the robotics work in Google's launch materials remains a set of demonstrations.

The big change

  • What changed: The model has moved from a restricted test to a stable, billed API release. Developers can now use a named endpoint and published limits to evaluate Argon in their own workflows; Google AI Pro subscribers get Argon in the Gemini app, but the Pro subscription does not include API or AI Studio access to this model.
  • Why it matters: The separate access paths affect both who can try Argon and what a test costs. A developer can budget tokens and choose an output limit, while an app subscriber can use the model without integrating an API. Google's robotics examples show possible uses but do not establish how reliably those systems work outside the reported demonstrations.
  • What to watch: Google's introductory API rates expire after November 30. Teams evaluating long outputs also need to decide whether background execution and its larger output budget fit their tasks; the default synchronous limit remains far below the shared million-token window.

API access, limits and price

Google's October 13 release note labels gemini-4-argon generally available. Its developer announcement says access is through the paid tier of the Gemini API. The model page lists a stable model ID, text output, and text, image, video, audio and PDF inputs. It lists function calling, code execution and search grounding among supported features; computer use is marked preview, and the Live API and image generation are not supported.

The model's 1,048,576-token context limit is shared across input, thinking and output. The model page lists 65,536 tokens as the default synchronous maximum output. Google's Argon guide says an Interactions API response can exceed 64K tokens through background execution: set background: true and max_output_tokens above 65,536, then collect text across the completed interaction's output steps. The total remains bounded by the shared window after input tokens are counted. The default thinking level is medium, with low and high also available. This is documented API behavior; BIG CHANGE has not run a long-output request.

The published introductory rates through November 30 are $2 per million input tokens and $10 per million output tokens, including thinking tokens. Google lists $0.10 per million cached input tokens on an implicit-cache hit. From December 1, it lists $4, $20 and $0.20 respectively. The guide says the API Free tier and Google AI Pro plan do not provide Argon access. Token use and any applicable service terms still determine the bill for a particular run.

Studio, Antigravity and the Gemini app

Google's developer announcement offers Argon in Google AI Studio and Google Antigravity to Google AI Ultra subscribers. Its robotics developer post repeats those entitlements and says Ultra subscribers can select gemini-4-argon in Antigravity. Those subscriptions are separate from the paid API tier.

The Gemini app announcement names both Google AI Pro and Ultra subscribers as eligible to use Argon and give Gemini tasks. It also describes an Antigravity coding interface in the Gemini macOS app. Thus a Pro subscriber's app access should not be read as access to Argon in AI Studio or the Gemini API. Google describes the model's app behavior and coding improvements, but this article does not treat those product claims as independent test results.

Robotics examples show a different scope of evidence

Google's robotics post shows Argon in research demonstrations, including a Boston Dynamics Spot task involving a foot-pedal bin and a MuJoCo simulation, and a robot-arm task using camera views and action primitives. Google also reports an 87% result on its embodied-reasoning benchmark. These are Google's descriptions and measurements. They do not amount to a new robotics API entitlement, a documented commercial deployment or an independently reproduced reliability result. The available developer path for trying Argon in robotics work is the same paid API or eligible AI Ultra developer products described above.

The October release resolves the main access question left by the September 30 restricted launch: developers now have a stable model ID, published prices and operational limits. Whether Argon is the right choice for a particular agent, coding task or robot still requires tests under that task's own conditions.

Sources & further reading

  • Gemini API release notes (October 13, 2026). Google's dated record identifies gemini-4-argon as generally available.
  • What's new in Gemini 4 Argon (updated October 13, 2026). Official API guide for the model ID, thinking levels, long-output workflow, introductory and later token rates, and API access exclusions. Prices and limits can change.
  • Gemini 4 Argon model page (checked October 14, 2026). Official input/output, context, default output and feature support table. A supported feature is not a measured result in BIG CHANGE's hands.
  • Gemini 4 Argon is now available for developers (October 13, 2026). Google's announcement distinguishes the paid API from AI Ultra access to Studio and Antigravity.
  • Gemini 4 Argon comes to the Gemini app (October 13, 2026). Google's app announcement names Pro and Ultra subscribers and describes the macOS coding interface. Behavior claims are Google's.
  • Gemini 4 Argon for Robotics (October 13, 2026). Google describes research demonstrations and its benchmark result, and repeats developer access paths. BIG CHANGE has not reproduced the demonstrations or evaluated robot reliability.