AI-300 Practice Question 90: Reasoning Effort Setting
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 9 · question 10 of 10
A team uses a reasoning model to sort support tickets into five simple categories. Latency and token cost are high, because the model produces long internal reasoning for a simple task. Which change reduces both most directly?
Show answer and explanation
Answer:
The reasoning_effort parameter controls how much a reasoning model thinks before answering. A higher effort generally produces more reasoning tokens and takes longer. For a simple classification task, a lower value reduces both latency and cost (option D).
- A is wrong because a higher effort makes the model think longer, which increases reasoning tokens, cost and latency. - B is wrong because a larger limit only allows more tokens to be spent and does not reduce the reasoning done. - C is wrong because temperature affects sampling randomness. It does not switch off or shorten the reasoning.
Reference: Azure OpenAI reasoning models