Gemini 4 Argon
Google's first Gemini 4 model and new frontier tier, built for long-horizon software engineering, enterprise knowledge work, and cybersecurity defense — released first to vetted cyber defenders through the Fairwind Program
Gemini 4 Argon
Google • September 2026
Training Data
Not disclosed
Gemini 4 Argon
September 2026
Parameters
Not disclosed
Training Method
Not disclosed
Context Window
Not disclosed
Knowledge Cutoff
Not disclosed
Key Features
1M-Token Output Limit • Autonomous Vulnerability Patching • Fairwind Program First
Capabilities
Coding: Outstanding
Cybersecurity: Outstanding
Long Video: Outstanding
What's New in This Version
77.9% on DeepSWE v1.1 (vs 65.3% for Gemini 3.7 Flash), #1 on AutomationBench at 51.3%, and 91.7% on LVBench, with output limits raised from 64K to 1M tokens; $2/$10 per 1M tokens introductory, then $4/$20
Google's first Gemini 4 model and new frontier tier, built for long-horizon software engineering, enterprise knowledge work, and cybersecurity defense — released first to vetted cyber defenders through the Fairwind Program
What's New in This Version
77.9% on DeepSWE v1.1 (vs 65.3% for Gemini 3.7 Flash), #1 on AutomationBench at 51.3%, and 91.7% on LVBench, with output limits raised from 64K to 1M tokens; $2/$10 per 1M tokens introductory, then $4/$20
Technical Specifications
Key Features
Capabilities
Other Google Models
Explore more models from Google
Gemini 3.8 Flash
Google's most intelligent workhorse model, taking extra reasoning steps and iterative tool calls on hard tasks for gains in software engineering, agentic work, and multi-step reasoning at the same price as Gemini 3.7 Flash
Gemini 3.8 Flash Cyber
A cybersecurity post-training of Gemini 3.8 Flash with more permissive safeguards, offered only to trusted defenders — government authorities, critical infrastructure operators, and software maintainers — through Google's Fairwind Program
Gemini 3.7 Flash
Google's most intelligent workhorse model for coding and agents, refining Gemini 3.6 Flash with algorithmic improvements to its core reasoning foundation
