The new model's opening weekend tripled Gemini 3's token volume, and Google says demand may outrun its supply
Google ’s newest AI model had a busy first weekend. Gemini 4 Argon processed nearly 25 trillion tokens in its first 72 hours of general availability.
That is roughly three times what Gemini 3 handled over the same stretch.
It also raises a less glamorous question. Google may not have enough computing capacity to keep up with the people who want to use Argon.
Start with tokens, since the whole milestone hinges on them. A token is a small chunk of text, often a word or a fragment of one, that a model reads or writes.
Every prompt a developer sends and every answer the model returns gets counted in tokens.
The rollout came in two stages. Google first announced Argon on September 30, 2026, with an initial limited launch aimed at trusted cybersecurity partners.
General availability followed on October 13, 2026. The 72-hour token count is measured from that wider release.
The headline feature is the context window. Argon supports a 1 million token input-output limit, a big jump from the 64K limit of Gemini 3.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Google is pitching Argon for sustained, multi-step work. The model is built for long-horizon tasks where it reasons through a problem over many steps with less hand-holding.
Target use cases include coding, general knowledge work and finance.
Google also pointed to internal results. The model reportedly lifted coding task submissions by over 35% while reducing rollbacks, meaning fewer changes had to be undone after the fact.
Developers accessing Argon through the API currently pay an introductory rate. That works out to $2 per million input tokens and $10 per million output tokens.
The discount has an expiration date. After November 30, 2026, those rates will double.
The gap between input and output pricing matters too. Output tokens cost five times as much as input tokens under the introductory scheme.
Google CEO Sundar Pichai has acknowledged ongoing capacity constraints amid strong demand for the new model.
Google also expects demand to outpace supply as Argon gains traction.
For enterprises, Argon’s 1 million token window opens up workflows that simply did not fit inside Gemini 3’s 64K limit. Teams working with large codebases, lengthy financial filings or sprawling internal documents may be able to hand the model the full picture rather than slicing it into pieces.
The phased rollout, starting with cybersecurity partners, also reflects a broader pattern of releasing powerful tools in stages rather than all at once.