Skip to main content
Smart search finds candidate models from a description of your task and explains the matches. Use filters, model details, and your own testing to make the final choice.

​
Describe your task

  1. Open the model catalog and describe your task in the search box.
  2. Include the input, required capabilities, and main constraints. For example: “Read product screenshots and return JSON, with low latency as a priority.”
  3. Submit the search or select a suggested task.
  4. Read the recommendation and open the candidates’ model details to check them.
You can select requirements such as image input or tool calling before submitting. Filters and smart search constrain the candidates together.
Describe the task without pasting API keys, customer information, or full business documents. If you already know a model’s name or ID, use regular search.

​
Understand match scores

A match score indicates how well a model fits this search, on a scale of 0–100. Use it to compare the current candidates. It is not a general model ranking or a guarantee that a task will succeed. Regular search results do not show match scores. A missing score is not replaced with zero. The recommendation explains why a candidate may fit; check its details for pricing and capabilities. Edit your description and submit it again to get new recommendations. Switching between list and table views, viewing favorites, or changing filters does not automatically run another smart search. Keep essential requirements first, then adjust preferences such as cost or context length. Test shortlisted models in Playground with representative inputs.

​
Empty results or unavailable search

Related: Choose a model · Model details and comparison Last updated: 2026-10-13