Google has made its latest frontier artificial-intelligence model, Gemini 4 Argon, generally available — and for buyers the headline is as much about price as performance, with an introductory rate that doubles after 30 November.
The model, announced in late September and opened up on 13 October, reaches developers and enterprises through the Gemini API and paying subscribers in the Gemini app and in AI Mode in Search, according to the company’s announcement by chief AI architect Koray Kavukcuoglu. Introductory pricing is set at two dollars per million input tokens and ten dollars per million output tokens — doubling to four and twenty dollars respectively from 1 December — with deep discounts on cached input for repeat workloads, according to published breakdowns of Google’s pricing.
A Price War at the Frontier
The timing puts Argon squarely in a three-way pricing contest. With flagship models from the three leading AI labs all available on standard paid API keys, procurement teams can now compare a two-and-ten-dollar list price from two vendors against a four-and-twenty-dollar tag from the third — a spread wide enough to force serious token-budget modelling before the introductory window closes. For marketing and technology teams running AI features at scale, model choice is becoming a line item rather than a pure capability decision, and the infrastructure bills behind it are climbing just as surging memory prices push hardware costs upward.
What Marketers Should Watch in Search
There is also a marketing story inside the rollout. Because Argon now powers AI Mode for paying subscribers, free users and paying users may see different AI-generated answers — and brands could appear differently across those tiers, with no measurement tools yet outlined for advertisers or publishers to track the gap. As platforms keep pushing AI-driven shopping and discovery surfaces, the question of who sees what — and who can verify it — is becoming a media-planning issue, not just an engineering one.
Enterprise teams face a parallel migration question: switching models is no longer a one-line change, with behaviour differences and new output limits to work through before the December price step-up. AI-driven marketing workflows are already pulling budgets in this direction, and the buyers who complete their evaluations before the introductory window closes will be the ones who actually understand their cost per answer.
