CALIFORNIA / RankWire.AI / – Google has unveiled Gemini 4 Argon, its newest flagship artificial intelligence system designed for sophisticated professional applications. Announced on Sept. 30, Argon is positioned as the leading model in the Gemini 4 series, capable of supporting intricate software development, financial analysis, legal research, and cybersecurity tasks. It also manages extended reasoning and execution sequences. Currently, access remains limited to select users, with cybersecurity defenders utilizing Argon through the Fairwind Program.
The model significantly increases the maximum output limit to 1 million tokens, a substantial jump from the prior 64,000-token cap. This enhancement enables the model to handle longer, more complex tasks without requiring multiple sessions. The introductory API pricing starts at $2 for every million input tokens, while output tokens cost $10 per million during the same period. Cached inputs benefit from a 95% discount. Future pricing is expected to rise to $4 for input and $20 for output tokens.
Numerous employees at Google are already employing Argon for tasks such as coding, research, and writing. The internal teams have tested the system on projects involving data center optimization and large-scale software migrations. One initiative used Argon agents to facilitate C and C++ migrations to Rust, while another focused on memory profiling across data centers. These efforts resulted in freeing over 300 tebibytes of memory, with further savings identified through ongoing analysis of the same systems.
Argon enhances capacity for extensive technical tasks
Google reported a 77.9% performance score for Argon on DeepSWE v1.1, a benchmark assessing prolonged software engineering performance. Additional results from the company cover areas such as finance, legal work, automation, and multimodal tasks. Argon was developed by Google DeepMind as part of the broader Gemini model family, integrating coding tools with long-context reasoning and multimodal capabilities. Its expanded output capacity is tailored for workflows that involve many interconnected steps before completing a task.
Cybersecurity remains a core aspect of the initial deployment. Argon can detect, validate, and patch vulnerabilities in approved security environments. Wiz is utilizing the model through its Scan for Good initiative, which targets security weaknesses in public infrastructure. Google reported a 68% score on CWE-bench v1, a benchmark dedicated to vulnerability remediation. Selected security teams can also access Argon without standard cyber guardrails when working on authorized security tasks.
Limited public access during phased introduction
Google has yet to announce a definitive date for widespread public access to Gemini 4 Argon. The company is implementing a phased rollout, collecting feedback from early adopters. It is also participating in a voluntary U.S. government process providing pre-release access to advanced AI models. Eventually, access will extend to developers, enterprises, and consumers. Priority will likely be given to paid API users and Google AI Ultra subscribers, though an official launch date has not been disclosed.
Google confirmed that it does not plan to release Gemini 3.5 Pro, which was previously anticipated before the Gemini 4 series. Instead, Argon now represents the latest flagship model optimized for demanding reasoning and professional tasks. Other Gemini models are still available for users with different needs regarding performance and pricing. For now, Gemini 4 Argon remains accessible mainly to trusted testers, cybersecurity partners, and early-access programs, with broader rollout yet to commence.