Google Announced Gemini 4 Argon: Flagship Model with 1M Token Output Limit, Initially for Select Users Only

Alphabet Inc. ($GOOGL)

Argon is designed for coding, enterprise tasks, and cybersecurity. Pricing per Google blog: $2 per 1M input and $10 per 1M output tokens, this is an introductory period, after which $4 and $20. Access is rolled out in stages: first trusted testers and cyber defenders, then paid API clients and Google AI Ultra subscribers. Pichai himself admitted that GPT-6 Astra is ahead in frontier coding and science, while Claude Opus 5.5 holds leadership in terminal coding. Just before the announcement, Bloomberg wrote that some Google employees doubt the model's coding capabilities. Google disputes this. Alphabet shares rose about 2.5% in after-hours trading.

Facts

According to Pichai, GPT-6 Astra surpasses Argon in frontier coding and scientific benchmarks, while Claude Opus 5.5 maintains leadership in terminal coding tests. [trad_fin, September 30, 2026]
Google announced Gemini 4 Argon in its blog: a frontier model for coding, enterprise work, and cybersecurity, rollout coming soon, first to testers and cyber defenders, then to paid API clients and Google AI Ultra.
Google's blog states that after the introductory period, the price will be $4 per 1M input and $20 per 1M output tokens. [blog.google, September 30, 2026]
On the eve, Bloomberg reported that some Google employees believe strong benchmarks do not always translate to real coding results, and the model is very large and expensive to run. Google disputes this. [@wallstengine, September 30, 2026]
Sam_Badawi citing Google: 77.9% on DeepSWE v1.1, 51.3% on AutomationBench, 91.7% on LVBench, 68% on CWE-bench v1. Output limit 1M tokens versus previous 64K, Argon freed more than 300 TiB of data center memory. [@Sam_Badawi, September 30, 2026]

Watches

When Argon becomes available to all developers and consumers: no public date.
Whether benchmarks will be confirmed by independent tests after access opens.

Post from Telegram channel trad_fin · Post from Telegram channel markettwits · Post from Telegram channel nwsmkr