CALIFORNIA / RankWire.AI / – Google has unveiled Gemini 4 Argon, its new flagship artificial intelligence model for complex professional workloads. The company announced the model on Sept. 30 as the foundation of its Gemini 4 generation. Argon targets software engineering, financial and legal knowledge work, and cybersecurity defense. Google said the model can sustain deeper reasoning across long, multi-step tasks. The company has begun providing access to a limited group of cybersecurity defenders through its Fairwind Program.
Gemini 4 Argon raises the maximum output limit to 1 million tokens, compared with 64,000 tokens previously. The larger limit allows the model to continue working through lengthy tasks within a single processing trajectory. Google set introductory API pricing at $2 per million input tokens and $10 per million output tokens. Cached input tokens receive a 95% discount from the input price. After the introductory period, prices will rise to $4 and $20, respectively.
Google said thousands of its employees already use Argon across specialized coding, research and writing tasks. Internal teams have also applied the model to data center memory optimization and large code migrations. One project used Argon agents on C and C++ migrations to Rust. Another applied agents to memory profiling across Google data centers. The company said completed optimizations freed more than 300 tebibytes of memory, with further savings identified through the same work.
Expanded capacity for complex professional work
Google reported a 77.9% score for Gemini 4 Argon on DeepSWE v1.1, which measures long-running software engineering performance. The company also highlighted results across finance, legal work and automation benchmarks. Argon supports multimodal reasoning alongside coding and enterprise tasks. Google DeepMind developed the model as part of the broader Gemini family. The company said the increased output capacity helps Argon manage extended workflows that require many consecutive reasoning and execution steps.
Cybersecurity forms another major part of the model’s initial deployment. Google said Argon can find, validate and patch software vulnerabilities in controlled defensive settings. Wiz is using Argon through its Scan for Good initiative, which focuses on identifying security exposures in public infrastructure. Google reported that Argon scored 68% on CWE-bench v1, a benchmark for vulnerability remediation. The company is giving selected defenders access without its standard cyber guardrails for approved defensive work.
Limited rollout precedes wider Gemini 4 access
Google has not announced a specific date for broad public availability of Gemini 4 Argon. The company said it is using a phased release process and collecting feedback from early testers. It also participates in a voluntary U.S. government process for pre-release model access. Google said later availability will include developers, enterprises and consumers. Its rollout plans begin with paid API customers and Google AI Ultra subscribers, although the company has not provided a fixed launch date for those groups.
Google also said it has no plans to release Gemini 3.5 Pro, which had previously been scheduled for June. The decision leaves Gemini 4 Argon as the company’s latest flagship model for demanding reasoning and professional workloads. Google continues to offer other Gemini models for different performance and cost requirements. Argon enters that lineup with its larger output capacity, expanded coding features and specialized cybersecurity functions. Its current availability remains focused on trusted testers and selected defensive security partners.