Google has announced the release of Gemini 4 Argon, its latest frontier AI model, available this week to a select group of users. This model is part of Google's ongoing efforts to enhance AI capabilities in enterprise environments, particularly in areas requiring complex, long-horizon workloads.
Gemini 4 Argon is designed to handle intricate tasks in software engineering, enterprise knowledge work, including legal and financial analysis, and cybersecurity. Currently, it is being rolled out to a limited set of organizations through Google's Fairwind Program, which complies with the US government's voluntary early access guidelines for testing and improving AI models' safety before a general release. The model is priced at $2 per million input tokens and $10 per million output tokens initially, with prices expected to double after the introductory period.
The standout feature of Argon is its expanded token capacity, now supporting up to 1 million tokens, which significantly increases its ability to perform reasoning and multi-step tasks. Evaluation results show Argon scoring 68.9% on the Vals Index and 77.9% on DeepSWE v1.1, surpassing some competitors, though it does not lead in every benchmark. For instance, it scores lower than Claude Opus 5.5 on PostTrainBench for ML engineering.
Target users of Gemini 4 Argon include enterprises and developers engaged in complex coding, reasoning, and multimodal tasks. Google has internally used Argon to identify and optimize memory usage in its data centers, projecting significant resource savings.
Work implications: Gemini 4 Argon can enhance roles in software engineering, cybersecurity, and enterprise analysis by automating complex tasks and improving efficiency in data management and vulnerability detection.
Originally reported by InfoWorld