Gemini 4 Argon: Google's Next Frontier Model (2026)

The short answer

Gemini 4 Argon is Google's first model in the Gemini 4 generation. Rather than a broad consumer launch, Google is taking a deliberately phased approach: give the model to vetted cybersecurity defenders first, gather feedback, tighten guardrails, then open it up to developers, enterprises, and consumers^[1].

The most concrete spec is the 1M-token output limit, up from 64K. That headroom lets the model "think" through hundreds of thousands of tokens in a single trajectory, which Google argues adds depth for long, multi-step problems^[1].

Pricing

Gemini 4 Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off the input price^[1]. Note that this is a preview-stage price; Google says broader availability comes "as soon as possible" after guardrail iterations^[1].

Coding and enterprise benchmarks

• DeepSWE v1.1: a new state of the art at 77.9% on real-world, long-horizon software engineering^[1].
• Vals Index: the leading model across finance, coding, legal, and tax work^[1].
• AutomationBench (Zapier): ranks #1 with a score of 51.3%^[1].
• LVBench: state of the art at 91.7% for long-video understanding^[1].

Cybersecurity defense

The headline capability is defensive security. Argon is trained to autonomously find, validate, and patch software vulnerabilities, and Google is releasing it to trusted defenders without cyber guardrails so they can use its full capability^[1]. On CWE-bench v1, which measures vulnerability remediation, Argon ties for first with a top score of 68%^[1].

Cloud-security firm Wiz is already using Argon through its Scan for Good initiative. In an early demonstration, the model uncovered a critical vulnerability exposing sensitive personal information across hospital software — a severe risk earlier frontier models had missed^[1].

Availability and rollout

Argon is rolling out now to a set of trusted cyber defenders through the Fairwind Program, while Google engages with the U.S. government's voluntary pre-release model-access process^[1][3]. After guardrail iteration, Google says it will open access to developers, enterprises, and consumers — "starting with paid API customers and Google AI Ultra subscribers"^[1].

Google is also reinforcing safeguards ahead of the wider release, including misuse and prompt-injection defenses, misalignment monitoring, and hardening of its sandboxed evaluation environments^[1].

What it means for you

• Not available to everyone yet. Unless you're in the Fairwind Program cohort, you can't access Argon today^[1].
• Security teams should watch closely. The defensive-cybersecurity focus — and Wiz's early results — suggests Argon is aimed at vulnerability discovery and remediation first^[1].
• Price-conscious users get a preview signal. At $2/$10 per million tokens, Argon's introductory pricing matches the top of Google's current Gemini line^[1].

Bottom line

Verdict: Gemini 4 Argon is a frontier-model announcement with an unusually narrow launch — vetted cyber defenders first, everyone else later. For now it's a signal of where Google is heading on long-horizon reasoning and defensive security, not a tool most readers can sign up for today.

Frequently asked questions

Can I use Gemini 4 Argon right now?

Only if you're part of the Fairwind Program cohort of trusted cyber defenders. A broader release to paid API customers and Google AI Ultra subscribers is coming later^[1].

What is the output token limit?

1 million tokens, up from the previous 64K limit^[1].

How much will Gemini 4 Argon cost?

The introductory API price is $2 per million input tokens and $10 per million output tokens, with cached input at 95% off^[1].