Breaking

Google Announces Gemini 4 Argon for Coding, Enterprise Work and Cyber Defense

September 30, 2026 at 16:09 EDT

Foundation Models
AI Agents
Security

Google announced Gemini 4 Argon on September 30, 2026, positioning the frontier model for long-horizon coding, enterprise knowledge work and cyber defense while limiting initial access to trusted cyber defenders. The company says broader access will follow after safety testing and feedback, with paid API customers and Google AI Ultra subscribers first in line.

Sep 30, 2026 | Google

Gemini 4 Argon targets long-horizon work

A frontier model for coding, enterprise knowledge work and cyber defense is rolling out first to trusted defenders while broader access waits on safety testing.

77.9%

DeepSWE v1.1

68%

CWE-bench v1

51.3%

AutomationBench

$2 / $10

Intro price per 1M input / output tokens

Output headroom jumps from 64K to 1M tokens

64K

Previous limit

Argon limit

The increase is roughly 16x, built for deeper single-run reasoning and long agent trajectories.

Release path

Trusted defenders->Fairwind access->Guardrail feedback->Paid users next

Early proof

Internal coding and research use
Wiz cyber-defense deployment
Operational memory and Rust migration examples

Limits to watch

No broad public testing yet
Independent workload validation still pending
Safety controls shape the rollout pace

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.