CALIFORNIA / RankWire.AI / – Google has unveiled Gemini 4 Argon, its latest flagship AI designed to support sophisticated professional tasks. The company announced Argon on Sept. 30 as the primary model within the Gemini 4 series. It is engineered to assist with complex software development, financial analysis, legal research, and cybersecurity operations. Additionally, the model is capable of managing extended reasoning sequences and execution steps. Currently, access remains limited, with selected cybersecurity defenders utilizing Argon through the Fairwind Program.
The new model raises the maximum token output to 1 million, a significant jump from the previous cap of 64,000 tokens. This expanded capacity enables the model to handle longer tasks without needing to split work across multiple sessions. Initial API pricing begins at $2 per million input tokens, with output tokens costing $10 per million in the same period. Inputs stored in cache qualify for a 95% discount. Future pricing is expected to increase to $4 for input tokens and $20 for output tokens.
Currently, thousands of employees within the company utilize Argon for coding, research, and writing assignments. Internal teams have tested the model on projects such as data center optimization and large-scale software migrations. One example involved using Argon agents to facilitate C and C++ to Rust migrations. Another focused on memory profiling across data centers, resulting in the release of over 300 tebibytes of freed memory. Additional savings were identified through ongoing analysis of these systems.
Argon broadens capacity for extensive technical tasks
Google reported a 77.9% score for Argon on DeepSWE v1.1, assessing its performance in long-term software engineering challenges. The company also shared results in finance, legal work, automation, and multimodal activities. Developed by Google DeepMind as part of the broader Gemini model ecosystem, Argon integrates coding tools with the ability to process long contexts and multimodal inputs. Its increased output capacity is tailored to tasks requiring multiple interconnected steps to reach completion.
Security remains a key focus during the initial deployment phase. Argon can detect, verify, and patch software vulnerabilities within authorized defensive environments. Wiz is utilizing the model via its Scan for Good initiative, which aims to identify security flaws in public infrastructure. Google reported a 68% score on CWE-bench v1, a benchmark centered on vulnerability remediation. Selected security teams can also operate Argon outside standard cyber guardrails for approved defensive security tasks.
Limited public access during phased rollout
No definitive launch date has been set for general public access to Gemini 4 Argon. Google is implementing a phased release approach, collecting feedback from early users. It is also participating in a voluntary U.S. government process that grants pre-release access to advanced AI models. Once fully available, the rollout will include developers, enterprise clients, and consumers. Priority access is expected for paid API subscribers and Google AI Ultra members, though an official launch date has not yet been announced.
The company has also confirmed that Gemini 3.5 Pro will not be released. This model was anticipated prior to the Gemini 4 series. Argon now serves as the flagship offering for demanding reasoning tasks and professional workloads. Other Gemini models continue to be accessible, catering to diverse performance and pricing preferences. For now, Gemini 4 Argon remains in the hands of trusted testers, cybersecurity partners, and early-access programs, with wider availability still upcoming.