Model access2 min read

Google opens Argon to developers and paid Gemini users

Argon’s October 13 rollout expands access beyond its initial cyber-defense program. Google keeps a separate path for approved users seeking fewer cyber restrictions.

Argon moves beyond its initial access program

Gemini 4 Argon became broadly available on October 13 through Google’s developer API and paid consumer products. The rollout extends access beyond the trusted cyber defenders involved in its earlier Fairwind release.[1]

Developers and enterprises can use the API through AI Studio and Google’s Gemini platform. Google AI Pro and Ultra subscribers get Argon in the Gemini app, while Google names Ultra access for Antigravity. The model’s original release and this wider rollout are separate dates.[1][3]

The app pairs the model with longer-running work

Google describes improvements in how Argon handles lengthy tasks: more direct responses, less repetition and clearer progress updates. It is also bringing Antigravity’s coding interface into the Gemini macOS app and letting users switch from conversation to tasks that continue while they are away.[3]

These are application features around the model. A developer calling the API receives the capabilities and limits of that API route; the consumer app’s connected services and subscription access do not automatically transfer to another host.

A shared context window and a longer decoding option

The current API documentation specifies a unified context window of 1,048,576 tokens, shared by input and output. Default synchronous output is 65,536 tokens. A long-decode option supports output up to the unified limit, with the input still consuming part of that same budget.[4]

Google lists Low, Medium and High thinking settings, alongside function calling, structured outputs, code execution and grounding tools. Those settings describe supported operation; they do not establish a benchmark score for every effort level or tool combination.[4]

At launch, Google advertised introductory token rates of $2 per million input tokens and $10 per million output tokens. Its pricing documentation now supplies the dates and route-specific conditions, including a scheduled transition after November 30. The selected API tariff is the appropriate place to check those conditions.[1][5]

A separate access decision for fewer cyber restrictions

Google reports that Argon solved all 12 end-to-end simulated attack scenarios in its cybersecurity evaluation, compared with five for the previous generation. That is a result from Google’s controlled testing, rather than a count of real-world intrusions.[2]

The company is expanding Fairwind while making the standard model broadly available. Approved defenders can seek access with fewer cyber guardrails through that program. Public access to the standard API does not itself grant access to that configuration.[1][2]

Sources & context

Based on the sources below, with LLMCurrent’s analysis. Original reporting dated .

  1. Google · Public Argon rollout (opens in a new tab)
  2. Google DeepMind · Argon safety and security (opens in a new tab)
  3. Google · Argon in the Gemini app (opens in a new tab)
  4. Google · Current Argon API specification (opens in a new tab)
  5. Google · Current API pricing (opens in a new tab)
About this article

Prepared with AI assistance from the cited material. It does not claim firsthand reporting or independent testing. Source claims are attributed; interpretation is identified in the text. Contact us about a correction.

Continue reading

Open modelsMeta, NVIDIA and Reflection release open weights for very different machinesAI agentsClaude AI helped map the sky—and sent a fake police tip