Reasoning

Control how much reasoning the model performs before answering using the reasoning_effort parameter. Higher effort produces more thorough answers at the cost of latency.

curl -X POST http://localhost:18645/v1/responses   -H 'Authorization: Bearer anything'   -H 'Content-Type: application/json'   -d '{
  "model": "grok-4.3",
  "input": [
    {"role": "user", "content": "Prove the irrationality of sqrt(2)."}
  ],
  "reasoning": {"effort": "high"}
}'

From the CLI, use the --reasoning flag:

progrok search --reasoning "complex query here"