CALIFORNIA / RankWire.AI / – Google has unveiled Gemini 4 Argon, its latest flagship artificial intelligence system tailored for sophisticated professional applications. The company announced Argon on Sept. 30 as the premier model within the Gemini 4 series. It is designed to support intricate software engineering, financial analysis, legal research, and cybersecurity tasks. Additionally, the model excels at managing extended sequences of reasoning and execution. Currently, early access remains restricted, with selected cybersecurity defenders utilizing Argon through the Fairwind Program.
Argon significantly raises the maximum output threshold to 1 million tokens, a substantial jump from the prior 64,000-token cap. This enhancement enables the model to handle longer and more complex assignments without needing to split work into multiple sessions. Initial API pricing begins at $2 per million input tokens, while output tokens cost $10 per million during the same period. Inputs stored in cache benefit from a 95% discount. Future pricing adjustments are expected to move to $4 for input and $20 for output tokens.
A vast number of employees within Google are already employing Argon for coding, research, and content creation. Internal teams have tested its capabilities on optimizing data center operations and migrating large-scale software systems. One notable project involved deploying Argon agents to facilitate C and C++ migrations to Rust. Another project concentrated on memory profiling across data centers, which ultimately freed more than 300 tebibytes of memory. Continued analysis of these systems has also uncovered additional savings opportunities.
Argon enhances capacity for extensive technical projects
According to Google, Argon achieved a 77.9% score on DeepSWE v1.1, an evaluation metric for long-duration software engineering tasks. The company also published results for its performance in finance, legal research, automation, and multimodal functions. Developed by Google DeepMind as part of the broader Gemini model ecosystem, Argon integrates coding tools with long-context reasoning and multimodal processing abilities. Its expanded output capacity is intended to support workflows requiring multiple interconnected steps before finalization.
Cybersecurity remains a key aspect of the initial deployment. Argon can detect, verify, and address software vulnerabilities within authorized defensive environments. Wiz is leveraging the model through its Scan for Good initiative, which targets security flaws in public infrastructure. Google reported a 68% success rate on CWE-bench v1, a benchmark focused on vulnerability remediation. Selected cybersecurity teams also have the option to use Argon without standard guardrails when working on approved security tasks.
Limited public access scheduled during phased rollout
Google has not yet announced a definitive date for widespread public availability of Gemini 4 Argon. The release process involves multiple phases, with feedback from early users informing subsequent steps. The company also participates in a voluntary process with the U.S. government, granting pre-release access to select advanced AI models. Eventually, the broader rollout will include developers, enterprise clients, and consumers. Priority access is expected to be granted first to paid API subscribers and Google AI Ultra members, although no specific launch date has been set.
Furthermore, Google clarified that it has no plans to release Gemini 3.5 Pro, which had been anticipated prior to the launch of Gemini 4. Argon now stands as the leading flagship model optimized for demanding reasoning and professional use cases. Other Gemini variants remain available to accommodate users with varying performance requirements and budget considerations. For now, Gemini 4 Argon remains accessible primarily to trusted testers, cybersecurity partners, and select early-access programs while the full public release is still forthcoming.