LEVEL 01 · LESSON 08
Model families and sizes
Choose a model by measured task fit, not size alone.
What it is
A model family is a related group of models. Variants may differ in size, training, supported inputs, reasoning behavior, and operational requirements.
Why it matters
The largest model is not always the best business choice. Quality, speed, cost, deployment options, and task fit all matter.
A business example
A service desk uses a compact model to categorize ordinary tickets and a stronger model for ambiguous cases. It measures whether the smaller model misses urgent issues before expanding its role.
How it actually works
Parameter count describes one aspect of a model’s scale, but architecture and training also affect performance. Some architectures activate only part of their parameters for each token. Benchmark candidate models on representative tasks, including difficult examples and required languages. Compare total system cost, including retries and human corrections, rather than relying on a model’s name or a general leaderboard.
A common misconception
Two models with similar parameter counts need not have similar capability, memory requirements, or response speed.
ONE SENTENCE TO REMEMBER
Choose a model by measured task fit, not size alone.
CHECK YOUR UNDERSTANDING
You are choosing between a large and a small model from the same family to triage customer emails. What is the best basis for the decision?
Show the answer
Answer: D.
Parameter count is one aspect of scale; architecture and training also shape capability, memory needs, and speed. Benchmark candidates on your own task, including difficult examples and required languages, and compare total system cost, including retries and corrections.
Take it into your next meeting
Which errors would make a faster, cheaper model unacceptable for your use case?
One more concept, understood.
Mark this lesson complete to keep your place.