AI model · by Google
Gemini 4 Argon
Gemini 4 Argon is Google's model for multi-step reasoning, math, visual coding, and document analysis, launched October 13, 2026.
- Released 13 Oct 2026
- 1M tokens context
- Closed weights
- Checked against Google's page 19 Oct 2026
What Gemini 4 Argon is for
Google says Argon is aimed at multi-step tasks, spatial reasoning, precise math, visual coding, and analysis of multimodal documents and longitudinal data. Since its October 13, 2026 launch, developers and enterprises worldwide can access it through the API in AI Studio and the Gemini platform. Google AI Ultra subscribers get it in Google Antigravity, while both Google AI Ultra and Pro subscribers can use it through the Gemini app and AI Mode in Search.
What Google says it does well
- Improvements in spatial reasoning, mathematical precision, and visual coding
- Follows style instructions better and communicates more directly with less hedging
- Deep multimodal document comprehension, longitudinal data analysis, and interactive analytical sandboxes while maintaining strict grounding
Where to use it
- Via the API in AI Studio and Gemini platform for developers and enterprises
- Google Antigravity for Google AI Ultra subscribers
- Gemini app and AI Mode in Search for Google AI Ultra and Pro subscribers