Google's Gemini 4 Argon Beats Opus 5.5 And GPT-6 Astra At Coding, But You Can't Use It Yet
0:00 / 0:49
News
Google's Gemini 4 Argon Beats Opus 5.5 And GPT-6 Astra At Coding, But You Can't Use It Yet
calendar_today Date:
schedule Duration: 0:49
visibility Views: 48
database
Summary Report
Gemini 4 Argon tops DeepSWE and AutomationBench with a 1M-token output limit, but is limited to cyber defenders for now.
- 01. DeepSWE 77.9% vs Opus 5.5 74.2% and Astra 74.1%; Opus still leads Terminal-Bench
- 02. 1M-token output; Fairwind Program first; $2/$10 introductory, then $4/$20
Google has unveiled Gemini 4 Argon, its new frontier model for coding, enterprise knowledge work and cybersecurity defence. On Google's numbers it scores 77.9% on DeepSWE, ahead of Claude Opus 5.5 at 74.2% and GPT-6 Astra at 74.1%, and is nearly nine points clear of Opus on Zapier's AutomationBench, though Opus still leads Terminal-Bench. It can write up to a million tokens in one response. It's rolling out first to cyber defenders through the Fairwind Program, at $2 in and $10 out per million tokens introductory pricing, rising to $4 and $20.
Meta Data
Company:
LLM:
Model:
