Google unveils flagship Gemini 4 Argon model
Google has announced its flagship Gemini 4 Argon neural network, capable of generating up to 1 million output tokens in a single run.
The new model is designed to handle long-horizon, multi-step tasks in software engineering, cybersecurity, as well as financial and legal analysis. The previous output limit stood at 64,000 tokens. According to Google, the expanded volume allows AI agents to execute complex workflows without splitting the work across separate queries.
Internally, Google is already using Gemini 4 Argon to optimize data centers, where its proposed solutions have freed up more than 300 TB (TiB) of RAM, with the potential to save up to 1 PB (PiB). In addition, the model's agents have been involved in optimizing quantum algorithms and porting system code from C/C++ to Rust, including over 800,000 lines of the Fuchsia operating system's Zircon kernel and 32,000 lines of SIMD code from the libgav1 video decoder.
A broad rollout of the product has been deferred: the system will initially be deployed to trusted cybersecurity professionals under the Fairwind program. Access will later be opened to Gemini API users and Google AI Ultra subscribers.
What it means
Base API pricing will be $2 per 1 million input tokens and $10 per 1 million output tokens, with a 95% discount on cached prompts. After the introductory period ends, prices will rise to $4 and $20, respectively.
Sources
- Kod.ru: Google представила Gemini 4 Argon: что нового media outlet
- iXBT.com: Google представила Gemini 4 Argon с миллионом выходных токенов media outlet
- Хабр: Gemini 4 Argon уже оптимизирует инфраструктуру Google и переписывает код media outlet
- 3DNews: Google вернулась в гонку ИИ с Gemini 4 Argon, которая слишком опасна для публики media outlet
This article was prepared with the help of AI and translated from the Russian original. Spotted an error — let us know.
Published 1 October 2026, 07:16 · No updates · Russian original




Comments on this story live under its post in our Telegram channel. Comment on Telegram