A custom prompt uses a model from Foundry's catalog. What should model selection validate?

1. Always choose by the largest parameter count
2. Required capabilities, supported deployment, quality, latency, and cost
3. Ignore input modality
4. Assume the model's name proves every output is correct