CALIFORNIA / RankWire.AI / – Google has announced the launch of Gemini 4 Argon, its newest flagship artificial intelligence system designed for sophisticated professional applications. The company revealed Argon on Sept. 30, positioning it as the primary model within the Gemini 4 series. This model is capable of supporting complex tasks such as software development, financial analysis, legal research, and cybersecurity operations. Additionally, it can manage extended sequences of reasoning and execution. Currently, early access remains restricted, with a select group of cybersecurity experts utilizing Argon through the Fairwind Program.
The Argon model increases the maximum output limit to 1 million tokens, a significant jump from the prior ceiling of 64,000 tokens, enabling it to perform longer and more intricate tasks without the need to split work across multiple sessions. Initial API pricing begins at $2 per million input tokens, while output tokens cost $10 per million during the same period. Cached inputs benefit from a 95% discount, and future pricing is expected to be adjusted 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 also tested the model on projects involving data center optimization and large-scale software migrations, such as using Argon agents to assist in the transition of C and C++ to Rust, as well as memory profiling across data centers. The completed optimization efforts have freed over 300 tebibytes of memory, with additional savings identified through ongoing analysis of the same systems.
Argon broadens its capacity for extended technical tasks
According to Google, Argon scored 77.9% on DeepSWE v1.1, a benchmark evaluating extended software engineering performance, and the company has also published results in areas such as finance, legal work, automation, and multimodal tasks. Developed by Google DeepMind as part of the wider Gemini model family, Argon combines coding tools with capabilities for long-context reasoning and multimodal processing, with its increased output capacity enabling it to undertake workflows requiring many interconnected steps before completion.
Cybersecurity remains a key focus during the initial rollout phase, with Argon capable of identifying, verifying, and patching software vulnerabilities within approved defensive environments. Through its Scan for Good initiative, Wiz is deploying the model to detect security weaknesses in public infrastructure. Google reported a 68% score on CWE-bench v1, a benchmark dedicated to vulnerability remediation, and select cybersecurity defenders can use Argon without standard security guardrails when working on approved security tasks.
Limited public availability expected during phased deployment
No specific date has been set by Google for widespread public access to Gemini 4 Argon, as the company prefers a phased release process that includes gathering feedback from early users. Additionally, Google participates in a voluntary U.S. government process that grants pre-release access to advanced AI models. The eventual rollout will include developers, enterprise users, and consumers, with paid API subscribers and Google AI Ultra members expected to gain early access; however, an exact launch date remains unspecified.
The company has also clarified that it does not plan to release Gemini 3.5 Pro, which was previously expected before the introduction of Gemini 4. Instead, Argon now serves as the flagship model catering to demanding reasoning and professional workloads. Other Gemini models continue to be available to users with different performance requirements and price points. For the moment, Gemini 4 Argon is primarily accessible to trusted testers, cybersecurity partners, and select early-access programs, with broader availability still pending.