Google unveils new Gemini 4 Argon AI model
On September 30, Google announced its new frontier AI model Gemini 4 Argon, initially available only to trusted cybersecurity defenders, designed to handle complex, long-horizon tasks in software engineering and enterprise work.
Google has announced Gemini 4 Argon, its newest frontier artificial intelligence model, which the company describes as the next stage in its most capable AI systems. According to Google's blog, the model is built to sustain deep reasoning across complex, long-horizon workflows — spanning real-world software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense.
Phased rollout
Argon is currently rolling out only to a set of trusted cyber defenders through Google's Fairwind Program. The company says safely releasing capabilities at this level requires a gradual approach, and it is participating in the U.S. government's voluntary process for pre-release model access. Broader availability for developers, enterprises and consumers will follow later.
The introductory pricing is set at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at a 95% discount.
Internal use at Google
According to the company, Argon is already powering internal workflows, with thousands of Google employees using it for coding and deeper research. Examples cited include quantum computing algorithm optimization, where the model beat a published baseline by 40% in one case, and fleet-wide memory optimization across Google's data centers, freeing over 300 TiB of memory with an estimated total potential of up to 1 petabyte. Argon agents are also being used for large-scale migrations of C/C++ codebases to Rust, including Fuchsia's Zircon kernel, spanning more than 800,000 lines of code.
Technical capabilities
The model's output token limit has been expanded to 1 million tokens, up from the previous 64,000, allowing it to reason more deeply in a single pass. Argon posts leading results on several industry benchmarks, including a 77.9% score on DeepSWE v1.1 for software engineering tasks, 51.3% on Zapier's AutomationBench for business process automation, and 91.7% on LVBench, which measures long-video understanding.
Cybersecurity focus
Google says Argon can autonomously find, validate and patch critical software vulnerabilities. For trusted defenders and internal Google teams, the model will be released without cybersecurity guardrails so they can access its full defensive capabilities. Cybersecurity firm Wiz is already using Argon as part of its Scan for Good initiative, which helps protect critical public infrastructure free of charge.
