noun
A base model is the most basic version of an AI system, typically offering core functionality without advanced features. It’s usually smaller, faster, and cheaper to run than premium models but may lack accuracy or capabilities like handling complex queries or generating high-resolution images.
Base models are often derived from larger, more sophisticated models by reducing their size or complexity. For example, a language model might be distilled to a smaller version that fits on a smartphone, sacrificing some nuance in responses for speed and efficiency.
For users, a base model means lower costs and faster performance but with trade-offs in quality. For developers, it’s a starting point for customization. For instance, a base image-generation model might cost $0.002 per image instead of $0.01 for a premium version, but the results may be less detailed.
A chatbot using a base model might answer simple questions quickly but struggle with nuanced or creative requests.
In other languages
- FR modèle de base
- ES modelo base
- DE Grundmodell
See also: fine-tuning · inference cost