Google DeepMind released Gemini 4 Argon to the public on Tuesday, making the newest frontier model available to developers, enterprises, and consumer subscribers simultaneously.
Developers can call Argon through the API in AI Studio and the Gemini developer platform. Subscribers to Google AI Ultra access it inside Antigravity. Both Ultra and Pro tier subscribers find it in the Gemini app and within AI Mode in Search.
Google frames Argon as a tool for intensive professional tasks rather than conversational use. The company reports that Argon leads on the Vals Index, a benchmark that covers finance, coding, legal, and tax challenges. Google also claims the model answers with more directness and less hedging compared to prior Gemini releases.
The model costs $2 for each million input tokens and $10 for each million output tokens. Cached input receives a 95 percent discount. Google describes the rate as introductory, leaving room for later adjustments. According to internal data, teams that used Argon in Antigravity raised their agentic code submissions by over 35 percent and dropped rollbacks by almost a third.
Google also demonstrated autonomous capabilities. Examples include a detailed temple complex sculpted in Blender via multi-agent critique loops, and a self-contained 3D ray-tracing engine written entirely in Python without external libraries. Argon arrives as frontier competition narrows to a handful of labs releasing models within weeks of one another. Google is wagering that discounted token pricing combined with tight developer tool integration will draw enterprise workloads onto its platform.
