Google unveils Gemini 4 Argon after months of delay
Argon, the first top-tier Gemini 4 model, can output up to 1 million tokens in one go. Google has dropped plans to ship Gemini 3.5 Pro.
What changed
Argon is larger than any model in Google’s Pro line and is aimed at long-running coding and enterprise knowledge work. Its output limit jumps from 64,000 tokens to 1 million. On DeepSWE v1.1, Google reports 77.9% for Argon against 74.2% for Claude Opus 5.5 and 74.1% for GPT-6 Astra.
Price and independent tests
API pricing is $2 per million input tokens and $10 per million output tokens, with 95% off cached input. Independent testing puts Argon level with GPT-6 Astra at 53 on a general intelligence index, but its hallucination rate on a knowledge test is 15% against 51% for Astra.
Timing
Argon rolls out first to Google AI Ultra subscribers and paid API customers. Gemini 3.5 Pro, which Sundar Pichai had promised for June, will not ship.