---
title: Private Models
description: >-
  Understand Private Models in Clear Ideas, including when to use private
  execution, how it differs from Provider Models, and how it fits governed Agent
  runs.
ogTitle: Private Models
ogDescription: >-
  Learn how Private Models provide a governed private execution lane for
  sensitive and repeatable AI work in Clear Ideas.
ogImage: /assets/images/og/ai-private-models.webp
navigation:
  icon: fasl fa-lock-keyhole
---

# Private Models

Private Models are a model lane for governed AI work that needs a stronger private execution posture, predictable operating costs, or tighter organization model policy.

They appear alongside Provider Models in Clear Ideas so users and administrators can make model choices deliberately. Private Models do not replace Provider Models; they give teams another controlled execution option.

## Private Models and Provider Models

Clear Ideas separates model choices into practical lanes:

- **Private Models** - Clear Ideas-managed private execution routes for sensitive, bounded, or repeatable work where privacy posture and cost predictability matter.
- **Provider Models** - models from external AI model providers, useful for premium reasoning, broad multimodal capability, specialized model behavior, or tasks that benefit from a specific provider.
- **Intelligent** - Clear Ideas selects an appropriate approved model based on the task and policy.

Private Models are especially useful for recurring Agent runs, structured extraction, review checks, benchmark runs, internal summaries, and high-volume document-heavy work.

Provider Models remain valuable for Agent design, debugging, ambiguous analysis, high-reasoning synthesis, complex multimodal work, and tasks that exceed the current private model envelope.

## Zero Data Retention Provider Paths

Some AI paths may support Zero Data Retention through a specific provider or model arrangement. Use Zero Data Retention language only when the selected model path and account arrangement support it.

Private model execution and Zero Data Retention provider paths are related privacy choices, but they are not the same claim. Check the model, policy, and provider arrangement before relying on a Zero Data Retention posture.

## Agent Execution

Private Models are often a good runtime lane for stable governed Agents. A common pattern is:

1. Design and debug the Agent with the best approved model for the authoring task.
2. Run routine, repeatable, or sensitive steps with Private Models when they are capable enough for the work.
3. Keep Provider Models available for steps that need additional reasoning or specialization.
4. Preserve source context, model choices, generated files, benchmarks, and run evidence.

This keeps the model strategy honest: Private Models are useful for governed execution, while Provider Models remain the right choice for some tasks.

## Billing and AI Credits

Private Models use AI Credits. Where private execution setup or access charges apply, they are tracked separately from token usage so account reporting can distinguish model tokens from private runtime setup.

See [AI Credits](/ai/ai-credits) and [AI Usage](/billing-and-payments/ai-usage).

## Administrator Controls

Organization policy can allow Private Models, Provider Models, both, or a narrower set of approved models. Administrators should treat model policy as a privacy and governance control, not only a billing control.

See [Model Policy](/ai/model-policy) and [Organization Policies Detailed](/organizations/organization-policies-detailed).

## Related Documentation

- [Models](/ai/models)
- [Model Policy](/ai/model-policy)
- [Agent Runs](/agents/runs)
- [Controlled Tools and Egress](/agents/controlled-tools-and-egress)
- [Governed AI Records](/ai/governed-ai-records)