Google announces Gemini 4 Argon flagship AI model for complex workflows

Alphabet's Google has introduced a flagship artificial intelligence model designed to anchor its newest generation of systems, marking a major push at the frontier of artificial intelligence research. The model, designated as Gemini 4 Argon, arrives after months of delays and positions itself directly against competing systems from rivals such as OpenAI and Anthropic. According to Yahoo Finance reporting, the new model is larger than previous top-tier Pro offerings and is built specifically to sustain deep reasoning across complex, long-horizon workflows.

The rollout strategy for Argon differs markedly from previous consumer-facing releases. Rather than opening the floodgates to the general public, MarketScale notes that Google is treating the launch more like critical enterprise infrastructure. The model is currently being distributed through the Fairwind Program to trusted cybersecurity defenders and government partners. This controlled approach follows a White House luncheon where tech executives met with President Donald Trump to address rising safety concerns surrounding artificial intelligence.

Media additions

Google Unveils Gemini 4 Argon, Deep-Reasoning AI Model For Code And Cybersecurity

Tulsee Doshi, Google's Gemini model product lead, told CNBC that starting the rollout in this phased manner provides greater confidence while placing a strong defensive tool directly into the hands of security practitioners. The Fairwind Program, which launched earlier in September, already encompasses more than 650 organizations across government, critical infrastructure, and security partners such as CrowdStrike, Palo Alto Networks, Snowflake, and Wiz.

Technical specifications released by Google highlight a substantial expansion in capability. Most notably, the output token limit has been increased to one million tokens, up from 64,000 tokens. According to official company Blog posts, this expanded headroom allows the system to think deeply and generate hundreds of thousands of tokens in a single trajectory, solving intricate problems in one continuous pass.

Internal deployments are already underway across Google. Thousands of employees are using the model for specialized coding, debugging, and research tasks. 9to5Google reports that Argon agents have analyzed fleet-wide profiling telemetry to autonomously apply memory optimizations across Google data centers, freeing up hundreds of terabytes of memory without the purchase of additional hardware. Furthermore, the model is assisting quantum computing researchers and aiding in large-scale code migrations, translating C and C++ codebases into Rust across projects ranging from modest libraries to the massive Fuchsia OS Zircon kernel.

Cybersecurity defense is a primary pillar of the new model's release. Google stated that Argon can autonomously find, validate, and patch critical software vulnerabilities. For trusted defenders and internal teams, the company is making the model available without standard cyber guardrails to maximize its utility. Security firm Wiz utilized Argon through its Scan for Good initiative, successfully uncovering a critical vulnerability exposing sensitive personal information across healthcare software used globally—a severe risk that previous frontier models had failed to detect, according to Unite.AI analysis.

Before making the technology generally available, Google is actively engaged in the U.S. Government's voluntary process for pre-release model access.

Pricing for the API has been established with an introductory rate. Developers will pay $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95 percent.

What to Watch Next

• Gradual expansion of the Fairwind Program beyond the initial cohort of 650 cybersecurity and government organizations.
• Completion of pre-release safety evaluations alongside the U.S. Government's voluntary oversight framework.
• Broader commercial rollout beginning with paid API customers and Google AI Ultra subscribers before reaching general developers and consumers.
• Updates regarding safety guardrail adjustments across misuse prevention, prompt injection resilience, and system sandboxing.

For more updates on corporate intelligence and enterprise software developments, consult our ongoing Business coverage.