Google’s Gemini 4 Argon is its most advanced AI model, launched September 30. Here is what makes it different.

It can produce up to one million tokens in a single response, enough for whole books. It scored 77.9 percent on the DeepSWE coding test.

On the Vals Index agent test it reached 68.9 percent. Google is first offering it through the Fairwind early-access programme.

Paid API and Ultra access come later. Pricing starts at $2 per million input tokens and $10 per million output tokens.

The model competes directly with OpenAI’s GPT-6 Astra. For users, it means longer documents and harder coding tasks handled in one go.

Google says full safety testing was done. Developers can apply for early access now.

Argon supports one million tokens of output per response. It scored 77.9 percent on DeepSWE and 68.9 percent on the Vals Index. Access starts through Google’s Fairwind programme. Pricing is $2 per million input tokens and $10 for output. It rivals OpenAI’s GPT-6 Astra.