← Back to Documentation
Models & Providers
Using AI Models
Help users choose the right models without exposing provider complexity.
Key Steps
- Use model aliases so users see stable names even when provider model IDs change.
- Make OpenAI, Anthropic, Gemini, and internal models visible through one model picker.
- Review cost, latency, and policy before making a model available to everyone.
Article Scope
This guide focuses on the operational steps needed to run Orchestris with team access, provider control, and predictable usage.
Visual Reference
Model picker behavior
- Users see the models their organization and role allow.
- Default models should be safe, cost-effective, and available across common workflows.
- Internal model names should be clear enough for business users to understand when to choose them.
Choosing defaults
- Use fast, lower-cost models for everyday summarization and drafting.
- Reserve higher-capability models for complex reasoning, coding, or analysis.
- Use private models when prompts must stay inside your controlled infrastructure.
Operational review
- Watch latency and cost after enabling a new model.
- Keep a fallback model available for critical workflows.
- Remove models that are deprecated, unreliable, or no longer approved.
Related Topics
Models & Providers
OpenAI Integration
Connect OpenAI models and expose approved options through the provider catalog.
Models & Providers
Anthropic Integration
Connect Anthropic models for teams that need strong long-context and reasoning workflows.
Models & Providers
Local Model Integration
Connect llama.cpp, OpenAI-compatible gateways, or other production-ready internal model endpoints.