=== How our Fair Usage Policy works ===

No Token Counting: You will never be charged an overage bill for context window size, input volume, or massive output completions.

Concurrence-Based Boundaries: To prevent malicious network exploitation or reseller abuse, we apply standard parallel request limits (e.g., maximum 20-40 concurrent active model streams running simultaneously on your token string).

The Goal: Build apps freely. You will only hit a temporary request delay if your platform operates like an enterprise-level commercial proxy server.