Google: Gemini 4 Argon and WeatherNext 3 Featured in September

Google detailed a series of announcements in September 2026, blending Gemini models, consumer tools, and programs for organizations. Beyond products, several projects aim at scientific and observational uses, from the human genome to atmospheric methane.

Scientific and Space Projects Scale Up

A test satellite has been launched as part of Project Suncatcher to evaluate the performance of Google’s TPUs in a real space environment. A model named MAPL-EMIT tracks methane leaks, developed in collaboration with NASA's Jet Propulsion Laboratory. In the field of biomedicine, AlphaGenome Atlas is presented as a predictive database covering every possible variant of the human genome. For weather forecasting, WeatherNext 3 integrates real-time satellite data and promises accurate precipitation forecasts. Google also states that its tools cover over 300 languages spoken by 7 billion people. The company further highlights climate observation uses from space and global methane monitoring.

Tools and Access: Connected Apps, Study Notebook, and Fairwind Program

The Connected Apps aggregate productivity and creation software within Gemini. Gemini Notebook offers customizable flashcards and animated video summaries for study reviews. A collaborative agent named CC in Google Labs aims to manage family logistics and shared calendars. For organizations, the 3.8 Flash models are announced as available through a new initiative, the Fairwind Program. On the client side, Gemini for Windows is offered, while Googlebook is described as a laptop designed around Gemini intelligence. Google also mentions the addition of means to directly link its applications within Gemini.

Gemini Range and Creation: Context from 1 Million Tokens to Lyria 3.5

Gemini 4 Argon features a context of 1 million tokens and is positioned as focused on complex problems, including cybersecurity. Google describes it as a cutting-edge model with advanced reasoning capabilities. Gemini 3.8 Live combines two natural language audio models, announced for everyday conversations and multi-step reasoning without latency. The 3.8 Flash and 3.8 Flash Cyber variants showcase advancements in reasoning and coding. Natural voice models have been launched, and music generation capabilities are integrated, with Lyria 3.5 presented as expressive for producing rich arrangements and personalized songs. These elements are part of a series of updates recently released.