CALIFORNIA / RankWire.AI / – Google unveiled Gemini 4 Argon, its newest leading artificial intelligence system designed for sophisticated professional tasks. The announcement was made on Sept. 30, positioning Argon as the forefront model within the Gemini 4 series. This AI supports advanced software development, financial analysis, legal research, and cybersecurity operations. It is also capable of managing extended reasoning and execution sequences. Access remains initially restricted, with selected cybersecurity experts utilizing Argon through the Fairwind Program.
The maximum token output for Argon has been increased to 1 million, representing a significant jump from the earlier 64,000-token limit. This enhancement enables the model to handle longer and more complex tasks without requiring work to be split into multiple sessions. The introductory API pricing begins at $2 per million input tokens, with output tokens costing $10 per million during the same period. Inputs stored in cache benefit from a 95% discount. Future pricing will adjust to $4 for input tokens and $20 for output tokens.
Many company employees are already using Argon for programming, research, and composing tasks. Internal teams have tested the model on projects such as data center optimization and large-scale software migrations. One initiative employed Argon agents to facilitate C and C++ migrations to Rust, while another focused on memory profiling across data centers. These optimization efforts resulted in freeing over 300 tebibytes of memory. Additional cost savings have been identified through ongoing system analysis.
Enhanced capabilities for complex technical tasks with Argon
On DeepSWE v1.1, a benchmark for extended software engineering performance, Argon achieved a score of 77.9%. The company also published results demonstrating its effectiveness in finance, legal work, automation, and multimodal tasks. Argon is part of Google DeepMind’s broader Gemini model ecosystem, which integrates coding tools with long-context reasoning and multimodal functionalities. Its increased output capacity allows it to support workflows requiring multiple interconnected steps before completing a task.
Cybersecurity remains a core focus during the initial launch phase. Argon can detect, verify, and address software vulnerabilities within authorized defensive environments. Through its Scan for Good initiative, Wiz employs the model to identify security flaws in public infrastructure. Google has reported a 68% score on CWE-bench v1, a benchmark dedicated to vulnerability remediation. Selected security defenders can also utilize Argon without standard cyber guardrails for approved defensive security operations.
Public access is still in limited testing stages during a phased rollout
Google has not yet announced a specific date for widespread public access to Gemini 4 Argon. The release will occur gradually, with feedback being collected from early users. The company also participates in a voluntary process with the U.S. government, which grants pre-release access to advanced AI models. Full availability is planned for developers, enterprise clients, and consumers, with paid API users and Google AI Ultra subscribers likely to gain access first, although no official launch date has been set.
Additionally, Google clarified that it has no plans to release Gemini 3.5 Pro, which was initially anticipated prior to the Gemini 4 series. Argon now serves as the flagship model for demanding reasoning tasks and professional workloads. Other Gemini variants continue to be available to meet different performance and pricing requirements. For the time being, Gemini 4 Argon is primarily accessible to trusted testers, cybersecurity partners, and early access programs, with broader deployment still pending.