Google 2015 logo (Wikimedia Commons, public domain)
Google unveils Gemini 4 Argon, an AI model built to hunt software bugs
The first Gemini 4 model targets cybersecurity, software engineering, and enterprise work, with access limited to vetted defenders for now
Google has launched Gemini 4 Argon, the first model in its Gemini 4 series and one aimed squarely at defensive cybersecurity. The company says it can autonomously discover, validate, and patch software vulnerabilities.
The launch landed on September 30, 2026. Beyond security, Google is pitching Argon for long-running, complicated workflows in software engineering and in knowledge-heavy fields like law and finance.
The numbers behind Argon
The headline spec is output capacity. Argon can produce up to 1 million tokens in a single response, compared with a 64K token output limit on earlier models.
On benchmarks, Argon scored 77.9% on DeepSWE v1.1, a test built around real-world software engineering tasks. OpenAI’s GPT-6 Astra scored 74.1% on the same test, while Anthropic’s Claude Opus 5.5 posted 74.2%.
Security is where Google is planting its flag. Argon achieved a top score of 68% on the vulnerability remediation portion of CWE-bench v1, a benchmark focused on fixing known categories of software weaknesses.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Google has also shared internal results. The model helped move more than 800,000 lines of code in the Zircon kernel, part of the Fuchsia OS, over to Rust. The same work saved 300 TiB of memory across Google’s data centers.
Who gets access, and what it costs
Argon is not open to everyone yet. It is available through Google’s Fairwind Program, which includes more than 650 vetted cyber defense organizations along with internal Google teams.
Participants include security heavyweights CrowdStrike and Palo Alto Networks. Wiz, one of the early deployers, used the model to uncover significant flaws in widely used healthcare software.
Broader availability is expected for paid API customers and subscribers to Google AI Ultra. Google has not tied that expansion to a specific date in the material available.
On pricing, Argon starts at an introductory rate of $2 per million input tokens and $10 per million output tokens. The roadmap allows those rates to rise to $4 and $20 respectively.