Base model vs assistant system
The distinction is a model and product boundary, not a claim that all assistants follow the same recipe.
| Question | Base model | Assistant system |
|---|---|---|
| Starting behavior | Broad learned continuation behavior | Behavior shaped for interaction and particular tasks |
| Training | Pretraining objective and data | Often additional demonstrations or preference optimization |
| Surrounding software | May be a simple model runner | May include messages, policies, retrieval, and tools |
| Information access | Only what its computation receives and has learned | May receive external evidence if the host fetches it |
| Failure modes | Can continue undesirable or false patterns | Still can produce unsupported claims and tool mistakes |
MAKE THE DISTINCTION CONCRETE
A chat interface is not proof of web access. External search exists only if the system supplies and executes that capability.
Follow the evidence
Read the connected lessons for detailed mechanisms, assumptions, and primary references.
01020304
← All comparisonsBase model versus assistant
The model that continues text and the product that answers your message are related, but they are not the same thing.
Supervised fine-tuning: demonstrations of behavior
Examples of good responses shift a pretrained model toward the behavior an application needs.
Prompt structure: what the model receives
A conversation becomes a structured sequence with roles, separators, instructions, and context.
Tool use: letting a system act
A model can propose a tool call; application code validates it, runs it, and returns the result.