Google Unveils Gemini 4 Argon AI Model

Google has officially introduced Gemini 4 Argon, the first model in its new Gemini 4 generation. Google says Argon is designed to handle complex, long-running tasks across software engineering, finance, legal work and cybersecurity. The model is initially being made available to a limited group of trusted cybersecurity partners rather than the general public.

One of Argon’s major upgrades is its expanded output capacity. Google says the model can generate up to 1 million tokens, compared with 64,000 in previous models. The larger context and output capability are intended to allow Argon to work through lengthy, multi-step problems in a single workflow. Google is also highlighting its performance in software engineering, where it reported a 77.9% score on the DeepSWE v1.1 benchmark.

Cybersecurity is another major focus of Gemini 4 Argon. Google says the model can autonomously identify, validate and patch software vulnerabilities and is being tested by trusted cyber defenders through its Fairwind programme. Google also says cybersecurity company Wiz has used Argon to identify a critical vulnerability affecting healthcare software. These claims come from Google’s own testing and demonstrations and have not been independently verified across all of the company’s benchmarks.

Google is also positioning Argon as a tool for professional knowledge work. The company says its employees are already using the model for debugging, large-scale code migrations, research and other engineering tasks. Google reported that Argon agents helped analyse data-centre telemetry and identify memory optimisations that could eventually free more than 300 tebibytes of memory, while other agents have been involved in migrating large C/C++ codebases to Rust.

Gemini 4 Argon will initially remain in limited testing as Google continues strengthening its safety systems. The company says it is working with the U.S. government’s voluntary pre-release model-access process before expanding availability. Google has announced introductory API pricing of $2 per million input tokens and $10 per million output tokens, with cached inputs receiving a 95% discount. Wider access is expected to begin with paid API customers and Google AI Ultra subscribers, although Google has not provided a specific public-release date.