Google unveils Gemini 4 Argon with access limited to cyber defenders
Google's September 30 announcement starts Argon access with trusted cyber defenders, lists introductory API pricing and gives no general release date.
Newer coverageGemini 4 Argon Goes Global: API, Antigravity and Gemini App Access Open
Storyline / Report 1 of 3
Gemini 4 Argon access
Full timeline 3 reports
- DevelopingGoogle unveils Gemini 4 Argon with access limited to cyber defenders (current report)
- DevelopingGemini 4 Next? Antigravity Users Lose Older Models as Argon Labels Surface
- DevelopingGemini 4 Argon Goes Global: API, Antigravity and Gemini App Access Open
Who can access Gemini 4 Argon?
October 13 update: Google has since opened a wider global rollout through paid API access and eligible Antigravity and Gemini subscriptions. The access limits below describe its September 30 preview, not current eligibility.
Google’s September 30 announcement says the initial rollout is to trusted cyber defenders through its Fairwind Program. It identifies paid API customers and Google AI Ultra subscribers as the starting groups for a later, broader release, but supplies no general release date. Rechecked October 12, that announcement does not establish that ordinary developers or subscribers already have access.
Announced pricing and output limit
Google lists introductory API pricing of $2 per million input tokens and $10 per million output tokens, with cached input priced at 95% off the input rate. A footnote lists $4 per million input tokens and $20 per million output tokens after the introductory period, without specifying when that period ends. These are announced prices, not evidence of general API availability.
The announced one-million-token ceiling applies to output length, not the input context window. For the announcement source and related coverage, see the Gemini 4 Argon background page.
What Google announced
Google introduced Gemini 4 Argon on September 30 as a frontier model for long-running software engineering, enterprise knowledge work and cybersecurity defense. Its first rollout is to trusted cyber defenders through Google’s Fairwind Program. The company says it is gathering feedback and strengthening safeguards before extending access to developers, enterprises and consumers. It names paid API customers and Google AI Ultra subscribers as later groups, but gives no date for general availability. The announcement is not a general release.
Google says the model’s maximum output length is one million tokens, up from a previous 64,000-token limit. This is an output ceiling, not a claim about the size of the input context window. The company reports strong results on coding and business-task evaluations and describes internal projects involving code migration and vulnerability remediation. Those figures and examples come from Google; the reviewed announcement does not establish how Argon performs in every outside workflow.
Why access is phased
Google says the model’s cyber capabilities require additional testing for misuse, prompt injection and agent actions that exceed a user’s intent. It plans to give selected defenders access without its normal cyber guardrails for defensive work, while broader access follows further safety work. This creates a useful distinction between an announced model and one that ordinary developers can test. Pricing is outlined in the announcement, but a wider launch date and independent performance evidence remain open.