Models & LLM Management

This section covers how Unique selects, validates, and manages the language models available on the platform. How We Evaluate Models describes the four-stage funnel used to assess models — from public benchmark screening and functional reliability testing through human-run financial validation and continuous automated regression. Benchmarking provides the framework customers can use to test models against their own use cases. Model Usage and Cost Management explains how administrators track token consumption and estimated spend per user, assistant, or application, and how to set per-user daily spending limits. Define Language Model Info via Environment Variable covers the environment variable approach for configuring which models are available in a given deployment.