Gemini 4 Argon opens to everyone – 13 days after the limited launch
On October 13, 2026, Google broadened its flagship model Gemini 4 Argon to Google AI Pro and Ultra subscribers, developers and enterprises — just 13 days after the model was launched exclusively for vetted cyber defenders. At the same time, Google is expanding the related Fairwind Program, which gives selected security partners a version of the model without cyber safety guardrails, to several hundred new partners. The company says the model underwent additional safety testing and a voluntary pre-release review at the White House; what that testing actually involved has not been made public.
The two-phase rollout
Argon was first announced on September 30 in a blog post by Google DeepMind SVP Koray Kavukcuoglu, according to Mashable. At that point, the model was restricted to "trusted cyber defenders" — vetted cybersecurity defenders — in Google's Fairwind Program.
On October 13, 13 days later, the model was made broadly available. Google AI Pro and Ultra subscribers can now use Argon in the Gemini app, and the model is available to developers and enterprises via API, according to Mashable and VentureBeat.
In the meantime, Google said it will open the Fairwind Program to several hundred additional partners. The program gives well-vetted cybersecurity defenders a version of Argon without its cyber guardrails, Reuters via Yahoo Finance reports. That means at the very moment Google opened the model to the general public, the company simultaneously expanded the most unrestricted access — a version from which the cyber safeguards have been removed — to a larger circle of security partners.
The safety process, as Google describes it
In the two weeks between the launch and the broad rollout, Google says it conducted additional safety testing and took part in a voluntary pre-release model access process with the U.S. government, according to VentureBeat.
CNET writes that Google says it engaged in the White House's voluntary review process for new models, and that Gemini 4 passed the inspection of the company's safety framework. CNET notes at the same time that the process is closed to the public — "the public has no insight into," as the reporting puts it. That means neither the content of the review nor the criteria behind the "passed" assessment can be independently verified.
It is worth keeping these points separate: It is confirmed that Argon was launched in two phases with a 13-day gap in between. But the characterization of phase one as a meaningful safety process rests on Google's own statements. What the "additional safety testing" entailed, what the cyber guardrails in the Fairwind version actually block, and what the government pre-release review evaluated are not described in the available reporting.
Gizmodo characterizes this type of phased launch — "part safety protocol, part self-promotion" — as a practice that has become standard for frontier AI labs amid widespread concern about misaligned AI. That is editorial analysis, not a fact about Google's process, but it places the rollout in a larger context: a short, controlled launch phase serves both safety purposes and marketing.
Pricing and availability
The launch API pricing starts at $2 per million input tokens and $10 per million output tokens, according to Reuters. VentureBeat highlights that this is a limited introductory price: the rates are set to double after November 30, 2026. CNET and Gizmodo confirm the figures. For enterprises considering building on Argon, this means the cost basis is in practice time-limited through November.
Among the early customers testing the model, Google names Deloitte, LexisNexis, Ambience Healthcare and Isomorphic Labs, according to Reuters. The companies themselves have made no public statements about their use in the available source material.
The context: delays and reorganization
The Argon launch does not come from a calm Google. The company announced in September that it no longer planned to launch Gemini 3.5 Pro — a model that CEO Sundar Pichai had originally said would arrive in June, Reuters writes. In addition, Google DeepMind underwent an extensive leadership reorganization after the lab's co-founder and CEO Demis Hassabis stepped aside, and several leaders of the Gemini model teams left the company.
Against that background, the short 13-day phase between announcement and broad launch is worth noting. A longer, better-documented testing phase would have provided more to assess; Argon got barely two weeks of restricted access before becoming available to all paying subscribers and API customers.
What remains unknown
Several central questions remain open after the available reporting:
- The contents of the safety testing. Google's statement about "additional safety testing" is not specified anywhere: neither scope, methods nor results are described.
- The guardrails in the Fairwind version. What the cyber guardrails do — which capabilities they block, and how — is not documented in the public sources. That means it is also unclear exactly what the several hundred new partners gain access to when they receive the version without them.
- The White House review. The process is voluntary and closed, and the "passed" assessment is Google's own characterization. There is no public documentation to measure it against.
- The scope of API access. Mashable notes that the announcement does not say whether API access has been expanded beyond Google's earlier rollout plans, even though Reuters and VentureBeat describe a broad launch to developers and enterprises.
All in all, the hard facts about Argon's launch are now in place: the dates, the pricing, the availability and the two-phase structure. But the very claim that the model is safe enough for everyone — built on two weeks of restricted testing and a closed government review — can for now only be verified by taking Google at its word.