Google (Gemini)
External APIs
Google serves Gemini 4 Argon at $2 and $10 per million until 30 November 2026, then double. The free tier trains the models.
Key points : Gemini 4 Argon · Promotional price until 30 November · Training on the free tier.
Models served : the range opens with Gemini 4 Argon, released on 13 October 2026, at $2 input and $10 output per million for a 1,048,576-token context, with a $0.10 cache. All three amounts double on 1 December 2026. Gemini 3.8 Flash remains served at $0.83 and $4.12.
Where the data goes
On unpaid offerings, including AI Studio and the free quota, Google uses submitted content to develop its products and machine learning technologies. Human reviewers may read inputs and outputs there. On the paid tier, data may be cached in any country where Google operates facilities.
Retention and training
The paid tier is mandatory to serve users in the European Economic Area, Switzerland or the United Kingdom. Tuning content is kept as long as the tuned model exists, and disappears with it. Google publishes no duration for other inputs.
When to use it from a sovereign platform
Value for money is good until 30 November, and that is precisely what should raise a flag. Any cost projection beyond it rests on a doubled price. Nothing sensitive should go through the free tier, where training and human review are announced.