Google DeepMind debuts Gemini 4 Argon for autonomous reasoning and agent teams

Google DeepMind has unveiled Gemini 4 Argon, its next-generation frontier model engineered for autonomous workflows and persistent agent teams. The model can manage long-horizon computational tasks across local sandboxes and web environments.

Argon features an expanded 2-million-token native multimodal context window and native code execution sandboxes. Introductory developer pricing is set at $4 per million input tokens and $18 per million output tokens via Google AI Studio and Vertex AI.

The release highlights DeepMind's push to move from passive conversational chat toward autonomous systems that can debug, deploy, and verify their own outputs in isolated developer containers.

What this means

• Gemini 4 Argon targets multi-agent orchestration and sustained reasoning.
• Active context reaches 2 million tokens with native cross-modal embeddings.
• DeepMind integrates direct sandbox code execution into the core API.