AI Provider vs Inference Provider
People think an AI provider is a firm that hosts models, and it is not. An inference provider runs models on its servers and sells the answers by the token. In the EU AI Act, a provider is whoever puts an AI system on the market under its own name. A host can be both, or neither.
AI Provider
The maker of an AI system or model under the EU AI Act. It carries the most duties.
An AI provider is the person or firm that builds an AI system or model and puts it on the market under its own name. That is the EU AI Act's word for the maker. It does not matter if the product is sold or given away free. The provider has the most duties in the law. For a high-risk system, it must test it, write the papers and keep watch after launch. A firm can also become one just by putting its own name on another firm's system.
Inference Provider
A firm that runs AI models for you. You pay each time you use one.
An inference provider is a company that runs AI models on its own chips and sells you the answers. You send a prompt to its API, and you get the reply back. You pay for the tokens or the seconds you use. Most providers host many open models, so you can use them without buying a graphics card. Some build their own chips to answer faster.
In one line
- AI Provider
- The maker of an AI system or model under the EU AI Act. It carries the most duties.
- Inference Provider
- A firm that runs AI models for you. You pay each time you use one.