For authorised nerds only
Google has revealed Gemini 4 Argon, its newest AI model, but access is very limited for now.
Google says the model is designed for difficult coding tasks, legal and finance work, and cybersecurity.
At first, only a small group of “trusted cyber defenders” will be able to use it while Google continues testing its safety.
The model is already being used inside Google for tasks such as moving large amounts of code.
Google also says Gemini 4 Argon performed better than models from OpenAI and Anthropic on several tests.
In brief:
Gemini 4 Argon is built for coding, business tasks and cybersecurity.
Access is limited while Google carries out more safety testing.
A wider release is expected once more safeguards are added.
Google said hands where I can see them
Before releasing it more widely, Google says it wants to improve protections against misuse, prompt attacks and unsafe behaviour.
The announcement came one day after OpenAI launched its Dots AI agent and GPT-6.1 Sol model.
And of course this lands one day after OpenAI’s big DevDay. The AI labs are now releasing models like divorced parents competing over who bought the better birthday present. -MG