AIF-C01 ยท GenAI strengths, limits and AWS services
Choosing a GenAI model
Model type, capabilities, performance, constraints, compliance, cost, latency and complexity when choosing a model for a business need.
Key points
- Start from the business need, then weigh: model type and modality (text, image, embeddings), capabilities (reasoning, languages, context window), performance on your own data, constraints (data residency, licence), compliance, cost per token, latency, and integration complexity.
- The largest model is rarely the best fit; the cheapest one that meets the quality bar usually is.
Practice questions
Which factor describes how long a user waits for a model's reply?
- Latency
- Context window
- Licence type
- Modality
Show answer
Answer: A. Latency
Latency is response time. Throughput is volume per time, context window is input size, and cost is price per use.
A company must keep customer data in the EU and use a model that handles Polish well. Which factors should drive the shortlist?
- Whether the model is available in an EU Region
- Measured quality on Polish-language tasks
- How well known the model provider's brand is
- The colour scheme of the model's console page
- The number of social media followers of the model
Show answer
Answer: A. Whether the model is available in an EU Region, B. Measured quality on Polish-language tasks
Regional availability/data residency and language capability are direct requirements here. Logo design and provider popularity do not address them.
True or false: the model with the most parameters is always the best choice for a business use case.
Show answer
Answer: False
False. Larger models usually cost more and respond slower; a smaller model that meets the quality bar is often the better fit.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
