Google announced Gemini 4 Argon on October 13, 2026, describing it as a frontier model built for complex problems in coding, finance, legal work, cybersecurity and science. The announcement reports a 1M unified input-output token limit and says the model ranks #1 on the Vals Index, which measures performance across several economic sectors. Google also reports nearly 2x fewer harmful-request violations and a drop in coding "shortcut" attempts to 2%, compared with earlier versions.
Access depends on how you use it. Consumers reach the model through the Gemini app and AI Mode in Search, but Google's announcement ties that access to a Google AI Ultra or Pro subscription — the free tier is not listed as included. Developers can call it through the API in AI Studio or Android Studio, and enterprises through Gemini for Business or the API. Google also mentions two limited programs: six months of free top-tier Antigravity access for researchers who apply, and Fairwind Program access for cybersecurity professionals without the usual safety guardrails.
If you plan to use the API, Google's introductory pricing is $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted 95%, through November 30, 2026. After that date the price rises to $4 and $20 respectively. Google says the model is available globally today.
Before you rely on it for real work: confirm your subscription actually includes Gemini 4 Argon rather than an older model, and test one result you can already verify — a document summary, a calculation, a piece of code — before trusting an unfamiliar answer. The safety improvements Google reports reduce certain risks, but they do not remove the need to check outputs yourself.
Source: Google, "Gemini 4 Argon: Built to solve your hardest problems" (October 13, 2026)