The platform for governed, auditable AI.
Validate, benchmark, and monitor every model in your enterprise — with compliance evidence built in from the first run.
start free — no credit card required
Serve the platform locally
$uvicorn openvals_api.main:app --port 8000- 0+
- Models under governance
- 0+
- Gate runs executed
- 0+
- Provider integrations
- 0+
- Compliance frameworks
Trusted by builders — works with every provider you already use
Platform
Everything needed to trust a model
From intake to production monitoring — one governed workspace.
Why OpenVals
A systematic way to ship trusted AI
Step 1 · EVALUATE
Which model fits the use case?
Score candidates on accuracy, safety, latency, and cost with a trade-off lens.
Models128
Browse all modelsGovernance core
Built for teams who ship production AI
Evidence stays yours
Self-host the gate engine and keep evaluation data on your own infrastructure.
Audit-ready by default
Every run emits a tamper-evident trail regulators and reviewers can read.
Built on standards
Controls map natively to EU AI Act, NIST AI RMF, ISO 42001, and SOC 2.
Vendor agnostic
Works with any model, framework, or provider. No lock-in.
Compliance
Our standards
Frameworks and controls your validation evidence maps to — out of the box.
6+ frameworks tracked and expanding
Resources
Get started in seconds
Stay in the loop with OpenVals
Gate releases, new frameworks, and AI governance insights — delivered to your inbox.
Validation Providers
Run gates against models hosted anywhere — OpenAI, Anthropic, Bedrock, Vertex, or your own endpoints.
Enterprise controls
RBAC personas, credential vaults, and audit trails wired into every run — no extra plumbing.