Breaking

Google Unleashes Gemini 4 Argon With 1M-Token Reasoning and Strongest Coding Benchmark Yet

September 30, 2026 at 16:09 EDT

Foundation Models
Software Dev & Coding

Google DeepMind has announced Gemini 4 Argon, its next-generation frontier model built for deep reasoning across extended workflows, introducing a 1 million token output capacity and claiming top performance on long-context software engineering benchmarks. The model, revealed on September 30, 2026, represents a significant leap in frontier AI capability, with particular strength in complex coding migrations, enterprise knowledge work, and cybersecurity vulnerability remediation.

September 30, 2026 · Google DeepMind

Gemini 4 Argon: Frontier Reasoning With 1M Token Output

Google announces its next-generation frontier model with extended reasoning capacity and state-of-the-art coding performance, built for complex long-horizon software engineering and enterprise knowledge work.

DeepSWE Coding

77.9%

State-of-the-art on long-horizon software tasks

Output Capacity

Token output limit (15× previous generation)

Security (CWE)

68%

Tied for first on vulnerability remediation

Model Comparison: DeepSWE v1.1 Benchmark

77.9%

Gemini 4 Argon

74.2%

Claude Opus 5.5

74.1%

GPT-6 Astra

Availability

Currently rolling out through Google's Fairwind Program to trusted cybersecurity defenders. Broader availability to developers and enterprises expected as soon as possible. Introductory pricing: $2 per million input tokens, $10 per million output tokens.

Continue reading

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