Google has announced Gemini 4 Argon, its new frontier AI model, which is rolling out to a set of trusted cyber defenders through the company’s Fairwind Program. Built to sustain deep reasoning across complex, long-horizon workflows, the model is designed to deliver strong performance across real-world software engineering, enterprise knowledge work such as legal and finance tasks, and cybersecurity defense.
Google said safely releasing frontier capabilities at this level requires a phased approach, and the company is actively engaged in the US government’s voluntary process for pre-release model access while gradually expanding availability. The company plans to continue gathering feedback from early testers as it refines safeguards before making Argon available more broadly to developers, enterprises, and consumers. Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at a 95 percent discount compared to the standard input token rate.
Gemini 4 Argon is already powering Google’s internal workflows, with the company reporting that thousands of employees have highlighted the model’s strengths in specialized coding tasks, deeper research, and writing quality. Google cited several internal use cases, including quantum computing researchers using Argon to optimize the resource requirements of subroutines that bottleneck important applications, beating a published baseline by 40 percent in one example. A team of Argon agents also analyzed fleet-wide profiling data to identify and apply memory optimizations across Google’s data centers, freeing up more than 300 TiB of memory, with total estimated savings ranging from 500 TiB to 1 PiB. Separately, Argon agents are being used to migrate C and C++ codebases to Rust across various Google systems, including a rewrite of the company’s open source libgav1 video decoding software that Google said now runs 2.7 times faster than a previous Rust port, while producing identical video output.
To support more complex use cases, Google is expanding Argon’s output token limit to 1 million tokens, up from the previous 64,000 token limit, giving the model more room to reason through difficult problems in a single response. Ahead of wider availability, Google said it is strengthening frontier safeguards across four areas: defending against misuse for cyber or CBRN related attacks, improving resilience against indirect prompt injection attacks, monitoring for signs of model misalignment during task execution, and hardening the sandboxed environments used to test the model safely. Google said Argon currently leads on Gray Swan’s Indirect Prompt Injection benchmark among models it has tested. The company said it remains grateful to its initial cohort of cyber defenders and trusted testers, whose feedback will inform further refinements before Argon becomes available to paid API customers and Google AI Ultra subscribers as a first step toward broader release.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
