This note summarizes the Vercel changelog entry about Gemini 4 Argon from Google on AI Gateway, and what it means if you call models from code.
How to call it
The changelog says to use google/gemini-4-argon as the model name. If you already route other models through AI Gateway, the change is limited to that identifier.
The change that matters most: thinking levels
According to the changelog, Gemini 4 Argon uses thinking levels (low, medium, and high) instead of thinking budgets. If your code tuned a numeric budget for other models, here you pick a level.
The level is set with reasoning in the AI SDK or reasoning_effort in Chat Completions.
Capabilities the source cites
The post says the model has a 1M token context window, accepts text, image, PDF, and video input, and supports tool calling.
Cost
The changelog states that pricing reflects provider pricing with no markup. That is Vercel’s statement; check current rates in your dashboard before budgeting.
What to do with this
Try the model at a low thinking level on a real task of yours and move up only if quality demands it. This note has no measurements of its own: we did not compare quality or latency.
