< Back to all clusters
[TECHNOLOGY] · United States, South Korea, Hong Kong SAR China, France, Japan · 75 sources

started · updated

Google launches Gemini 4 Argon AI model for complex reasoning

Google has launched Gemini 4 Argon, a new frontier AI model designed for complex, multi-step tasks in fields such as software engineering, finance, law, cybersecurity, and scientific research. The model features a unified context window of up to one million tokens for input, thinking, and output.

Argon is available to developers and enterprises via the Gemini API, Google AI Studio, and Google Antigravity. Google AI Pro and Ultra subscribers can access the model through the Gemini app and Google Search's AI mode. For developers using the API, Google is offering an introductory price of $2 per million input tokens and $10 per million output tokens, though these rates are scheduled to double after November 30.

Internal applications have demonstrated significant efficiency gains. Google reported that using Argon within the Antigravity platform increased agentic code submissions by over 35% and reduced code rollbacks by approximately one-third. The Chrome team utilized the model to complete a major architectural refactoring of the Chromium codebase in weeks, a task previously estimated to take months. Additionally, the model has been used in scientific research to assist with quantum computing and life sciences, and in cybersecurity through the expanded Fairwind Program.

Entities

Alphabet · Anthropic · Antigravity · DeepMind · Gemini 4 Argon · Google · Google Cloud · Google DeepMind · OpenAI · Sundar Pichai · Thomas Kurian

Claims

What the coverage asserts, and how many sources carry each claim.

Sources