Question

Your team is building a customer-support summarization feature in Azure AI Foundry. The feature must summarize long support transcripts (up to 30,000 tokens) with high accuracy, but the product owner also requires that per-request cost stays low because millions of summaries are generated monthly. Before committing to a model, you want to make an evidence-based selection that balances quality against cost and context length. Which approach in Foundry should you use to make this decision?

Accepted Answer

Use the Foundry model catalog's model benchmarks and comparison tools to evaluate candidate models on quality, context window, and cost, then run your own evaluation on representative transcripts.

Answer

Deploy the highest-parameter frontier LLM available in the catalog, since larger models always produce the best summaries.

Answer

Select any small language model, because SLMs are always cheaper and sufficient for summarization tasks regardless of transcript length.

Answer

Pick the model with the lowest per-token price in the pricing page and skip evaluation to accelerate delivery.