CALIFORNIA / RankWire.AI / – Google has introduced Gemini 4 Argon, its new flagship artificial intelligence model for advanced professional tasks. The company announced Argon on Sept. 30 as the lead model in the Gemini 4 generation. It supports complex software development, financial analysis, legal research and cybersecurity work. The model also handles long sequences of reasoning and execution. Initial access remains limited, with selected cybersecurity defenders using Argon through the Fairwind Program.
Argon expands the maximum output limit to 1 million tokens, a sharp increase from the previous 64,000-token ceiling. That capacity allows the model to complete longer tasks without breaking work into as many separate sessions. Introductory API pricing starts at $2 per million input tokens. Output tokens cost $10 per million during the same period. Cached inputs receive a 95% discount. Prices will later move to $4 for input and $20 for output tokens.
Thousands of company employees already use Argon for coding, research and writing assignments. Internal teams have tested the model on data center optimization and large-scale software migration projects. One project used Argon agents to support C and C++ migrations to Rust. Another focused on memory profiling across data centers. The completed optimization work freed more than 300 tebibytes of memory. Teams also identified additional savings through continued analysis of the same systems.
Argon expands support for long technical tasks
Google reported a 77.9% result for Argon on DeepSWE v1.1, which measures extended software engineering performance. The company also released results covering finance, legal work, automation and multimodal tasks. Google DeepMind developed Argon as part of the broader Gemini model family. The model combines coding tools with long-context reasoning and multimodal capabilities. Its larger output capacity supports work that requires many connected steps before a task reaches completion.
Cybersecurity remains a central part of the initial rollout. Argon can identify, confirm and patch software vulnerabilities in approved defensive environments. Wiz is using 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 focused on vulnerability remediation. Selected defenders can also use Argon without standard cyber guardrails when working on approved defensive security tasks.
Public access remains limited during phased rollout
Google has not set a firm date for broad public access to Gemini 4 Argon. The company is using a phased release and gathering feedback from early users. It also participates in a voluntary U.S. government process that provides pre-release access to advanced AI models. Later availability will cover developers, enterprise customers and consumers. Paid API users and Google AI Ultra subscribers are expected to receive access first, though no fixed launch date has been announced.
The company also confirmed that it does not plan to release Gemini 3.5 Pro. That model had previously been expected before the Gemini 4 generation. Argon now serves as the latest flagship offering for demanding reasoning and professional workloads. Other Gemini models remain available for users with different performance and pricing needs. For now, Gemini 4 Argon remains focused on trusted testers, cybersecurity partners and limited early-access programs while broader availability has yet to begin.