Model
A model is the trained mathematical structure that maps inputs to outputs based on learned patterns.
What Is a Model?
A model is the artifact that results from training: a fixed set of parameters, arranged in a particular structure, that takes an input and produces an output. Everything a system "knows" after training is encoded in those parameters — the model itself has no separate memory or reasoning process outside of that structure.
Training Data
Learning SourceWhat the training process learns patterns from.
Training Process
One-Time StepRuns once, ahead of any real use.
Model (Parameters)
Fixed ArtifactEverything learned, encoded as fixed numbers.
New Input
Not Seen BeforeData the model wasn't trained on directly.
Output
No Extra MemoryProduced purely from the fixed parameters.
Key Idea
A model is a static artifact once training finishes — using it (inference) doesn't change it, only training does.
Common Mistakes
Assuming a model updates itself from use
A deployed model doesn't learn from the requests it handles unless it's explicitly retrained or fine-tuned on new data.
Treating "model" and "AI system" as interchangeable
A production AI system typically wraps a model with retrieval, tools, validation, and other application logic — the model is one component, not the whole system.
Assuming a bigger model is always a better choice
Model size is one factor among latency, cost, and task fit — a smaller model is often the better engineering choice for a simple task.
Interview Question
What is a model, and how is it different from training or inference?
A model is the trained artifact — a fixed structure of parameters — that maps inputs to outputs based on patterns learned during training. Training is the process that produces the model by adjusting its parameters against a dataset; inference is using that already-trained, unchanging model to get a prediction on new input. The model itself doesn't learn or change from being used — it stays static until it's retrained or fine-tuned.
What an interviewer may ask next
- Does a deployed model learn from the requests it handles?
- What's the difference between a model and an AI system built around it?
- Why isn't a bigger model always the right choice for a given task?
Explain It in 30 Seconds
A model is the trained artifact that results from training — a fixed structure of parameters that maps inputs to outputs based on learned patterns. It's static once training finishes: using it through inference doesn't change it, only retraining or fine-tuning does. A model is typically one component within a larger AI system that also includes retrieval, tools, and validation logic.