Launch Week 02 wrapped — explore all five launches
Confident AI
Docs
Pricing
Products
LLM Evaluation
LLM Evaluation

Benchmark LLM systems with research-backed metrics.

LLM Observability
LLM Observability

Trace, monitor, and alert on production LLM systems.

AI Red Teaming
AI Red Teaming

Stress-test LLM apps against adversarial attacks.

AI Governance
AI Governance

Enforce AI standards and controls across teams.

Book a DemoSign Up
Book a DemoSign Up

Pre-deployment Eval Controls

Gate deployments on a qualifying evaluation test run.

Included on the Enterprise plan. Book a demo, opens in a new tab. Not included on the Team plan. Not included on the Starter plan. Not included on the Free plan.

Pre-deployment eval controls gate deployment on a recent evaluation test run. Use them to require evidence that an AI application was evaluated under the conditions your organization expects before it is released.

When should you use pre-deployment eval controls?

  • When a release must be backed by a recent, qualifying evaluation test run.
  • When you need to verify that the approved model, prompt, dataset, or hyperparameters were tested.
  • When deployment should be blocked unless the intended application configuration has evaluation evidence.

Select the gating test run

Choose one of two selection methods:

Latest official run

The control selects the most recent test run marked official in the project. Use this method when your team explicitly promotes one run as the source of truth for release decisions.

Identifier and window

The control selects the latest completed test run that:

  • matches the configured identifier, and
  • completed within a rolling window of 7, 14, 30, or 90 days.

Use this method when a stable identifier represents a release pipeline or evaluation suite and the result must stay current.

Apply filters

You can add filters, including hyperparameter filters, to define which test run qualifies. The selected run must match every configured filter.

For example, filters can require the gating run to use a particular model, prompt version, dataset, or application configuration captured in its hyperparameters.

Example requirements

  • Require the latest official evaluation to use the approved production model.
  • Require a completed release-candidate evaluation from the last seven days.
  • Require the selected run to match the approved dataset and prompt version.
Scaling beyond prototype?For teams evaluating Confident AI in productionTalk to us

Last updated on

Built byConfident AI