How do I set thinking effort in the Leiolai API?
Leiolai API requests support low, medium, high, and xhigh effort, with low as the default. Use the effort field for your chosen request format. The model ID stays leiolai-1, and the price depends on the mode and effort selected for the request.
Read the Leiolai API reference
Which effort levels are available?
Leiolai accepts low, medium, high, and xhigh. A request without an effort value uses low.
Which field does Chat Completions use?
Leiolai Chat Completions uses reasoning_effort at the top level of the request. For example, set reasoning_effort to high alongside model and messages. The text Completions endpoint uses the same top-level effort field.
Chat Completions fieldsWhich field does Responses use?
Leiolai Responses uses reasoning.effort. Send a reasoning object with effort set to the desired value. This differs from Leiolai Chat Completions’ top-level reasoning_effort, so adapt the field when moving a request between formats.
Responses request formatWhich field does Messages use?
Leiolai’s Anthropic Messages format uses output_config.effort. Keep the Messages max_tokens field and conversation messages as usual, then set effort in output_config.
Messages effort and output settingsIs effort the same as an output token cap?
No. Effort selects Leiolai’s thinking depth. An output cap sets the maximum length of a finite answer: max_completion_tokens in Chat Completions, max_output_tokens in Responses, or max_tokens in Messages and text Completions.
Context, thinking, and output lengthDo I need a different model name for each effort?
No. Use leiolai-1 at every effort. Do not invent model IDs such as leiolai-1-high; configure the supported field on the same model request.
Use the current model IDHow should I compare the cost of different efforts?
Use the input and output rates for the Leiolai mode and effort you want to compare. Apply the same estimated token counts to each rate pair. For a completed request, use the returned usage counts when available.
Estimate costs at each effort Read metered token usageWhat should I do if my client has no effort selector?
A Leiolai request with no effort field uses low. If your client supports extra request-body fields, add reasoning_effort for Chat Completions, reasoning.effort for Responses, or output_config.effort for Messages. Keep the model ID as leiolai-1.