Skip to main content
Reasoning models such as deepseek-r1 and deepseek-v4.1-flash produce an internal chain-of-thought before emitting their final answers. Neosantara surfaces this thinking process in a dedicated reasoning_content field on /v1/chat/completions.

Billing Rules for Reasoning Tokens

Neosantara adheres to strict billing integrity principles:
  • Delivered Output Is Billable: Reasoning tokens delivered to the client (via reasoning_content) are counted as output tokens in usage.completion_tokens.
  • Usage Breakdown: The usage payload includes completion_tokens_details.reasoning_tokens allowing exact analysis of thinking token volume.
  • Model Integrity: Billing is calculated at the exact public rate of the model requested.

Available Reasoning Models

Next Steps