Select a small language model whose benchmark quality meets the task.

Deploy the largest available reasoning model.

Use Azure Translator for classification.

Use an image generation model.