Count Tokens in Messages (LLM API)

Use this endpoint when you need to check token usage before committing to an inference call — ideal for context window management, cost estimation, or prompt optimization. It is free, so it should be called proactively whenever there is uncertainty about whether a prompt fits within a model's context window. Prefer this over estimating tokens client-side, as tokenization rules differ per model and this endpoint uses the exact same tokenizer as the inference endpoint.

• Invalid model identifier returns a validation error — must be one of the supported enum values
• Malformed messages array (missing role or content fields) causes a 400 error
• Empty messages array may return zero or trigger a validation error
• Model-specific tokenization differences mean counts are model-specific and not interchangeable across models