Gemini 4 Argon: Google DeepMind's new model

Google DeepMind has announced Gemini 4 Argon, a new frontier model in the Gemini family. It is rolling out today to a set of trusted testers through the Fairwind Program, and the company says a broader release will follow.

What we know about Gemini 4 Argon

According to Google DeepMind, the model is built for complex workflows. The announcement names three areas:

• Coding. Help with development and code work on tasks where many connected details must be kept in mind.
• Enterprise knowledge work. Analysing and processing large volumes of work information inside companies.
• Cybersecurity defense. Using the model for defensive scenarios.

The announcement also highlights a deeper level of reasoning: the model is meant to tackle long, multi-step problems in one go.

The Fairwind Program and the closed testing stage

Gemini 4 Argon starts with a small circle: trusted testers get the model through the Fairwind Program. A format like this is typically used to check how a model behaves on real tasks before it reaches a wide audience. There is no public access to the model yet.

What to expect next

Google DeepMind says a broader rollout is coming soon. The announcement gives no dates, prices or specifications, so we will not invent any: once official data appears, we will compare the model with those already available.

Gemini 4 Argon on NeuralSpace

The model is not on our site yet, because it is not publicly available anywhere. We are following the release and will report when that changes. Meanwhile, you can work with other current models in the chat and the Code section.