CALIFORNIA / RankWire.AI / – Google revealed Gemini 4 Argon, its latest top-tier artificial intelligence designed for sophisticated professional tasks, during an announcement on September 30. This model forms the core of the upcoming Gemini 4 generation, with a focus on applications in software engineering, finance, legal analysis, and cybersecurity. Google emphasized that Argon can handle more profound reasoning over extended, multi-step processes. The company has already begun granting select cybersecurity professionals access via its Fairwind Program.
Compared to previous iterations, Gemini 4 Argon increases the maximum token output to 1 million, surpassing the earlier limit of 64,000 tokens. This enhancement enables the model to process and complete lengthy tasks within a single sequence. The initial API pricing has been set at $2 per million input tokens and $10 per million output tokens, with cached input tokens enjoying a 95% discount. Following the initial rollout, prices are scheduled to increase to $4 and $20, respectively.
Google stated that thousands of its employees are already utilizing Argon for specialized coding, research, and writing tasks. Internal teams have also employed the model for optimizing data center memory and migrating large codebases. One project involved using Argon agents to convert C and C++ code to Rust, while another used agents for memory profiling across Google’s data centers. Google reported that these efforts resulted in freeing more than 300 tebibytes of memory, with additional savings identified through ongoing work.
Enhanced capabilities for intricate professional applications
On DeepSWE v1.1, which assesses performance in extensive software engineering tasks, Google assigned Gemini 4 Argon a score of 77.9%. The company also highlighted positive results in financial, legal, and automation benchmarks. Argon supports multimodal reasoning along with coding and enterprise workflows. Developed by Google DeepMind as part of the broader Gemini series, the model’s increased output capacity is designed to manage long, complex workflows that involve numerous reasoning and execution stages.
Another key aspect of the initial deployment involves cybersecurity. Google mentioned that Argon is capable of detecting, confirming, and fixing software vulnerabilities in controlled defensive environments. Through its Scan for Good initiative, Wiz is utilizing Argon to identify security gaps in public infrastructure. The model scored 68% on CWE-bench v1, a benchmark for vulnerability remediation. Google is providing selected cybersecurity defenders with access to Argon without its standard cyber guardrails for approved protective tasks.
Early limited release before broader availability of Gemini 4
There has been no official announcement regarding a specific date for the wider public release of Gemini 4 Argon. Google explained that it is employing a phased rollout approach and gathering feedback from early testers. The company also participates in a voluntary pre-release model access process with the U.S. government. In the future, the model will be available to developers, enterprises, and consumers. The initial rollout will target paid API users and Google AI Ultra subscribers, though no fixed launch date has been specified for these groups.
Furthermore, Google clarified that it has no plans to release Gemini 3.5 Pro, which was previously scheduled for June. This decision positions Gemini 4 Argon as the company’s newest flagship for demanding reasoning and professional workloads. While other Gemini models remain available for different performance and budget needs, Argon distinguishes itself with its larger output capacity, expanded coding capabilities, and dedicated cybersecurity features. Currently, the model is accessible only to trusted testers and select security partners involved in defensive operations.