Start free. Scale to the audit.
Plans for a single engineer trying evals on a Friday afternoon, and for a regulated enterprise that needs everything in its own cloud with a signed DPA.
For individuals and prototypes.
- 25,000 graded cases / month
- Full grader library
- CLI, Python and TypeScript SDKs
- Community support
- 7-day result history
For product teams shipping to production.
- 1M graded cases / month
- CI gates and regression diffs
- Online evals on production traces
- Guard: 5M decisions / month
- Nightly automated red teaming
- SSO, 1-year history, Slack support
For regulated and high-scale deployments.
- Unlimited suites and volume tiers
- VPC, single-tenant or air-gapped
- Guard at any scale, 99.99% SLA
- Expert red-team engagements
- Custom safety data and gold sets
- EU AI Act and NIST AI RMF evidence packs
- Named research engineer
Engagements and data.
| Service | Typical scope | Duration | From |
|---|---|---|---|
| Pilot: release-gating suite | One use case, threat model, suite, baseline | 30 days | $25,000 |
| Pre-launch red-team assessment | One agent or application, retest included | 2–3 weeks | $48,000 |
| Frontier model evaluation | Capability and misuse evals with domain experts | 4–8 weeks | Scoped |
| Custom safety dataset | Red-team corpus or preference pairs to your policy | 3–10 weeks | Scoped |
What's in each plan.
| Capability | Developer | Team | Enterprise |
|---|---|---|---|
| Grader library (1,400) | ✓ | ✓ | ✓ |
| Custom graders and rubrics | ✓ | ✓ | ✓ |
| CI gates with statistical tests | – | ✓ | ✓ |
| Agent trajectory evaluation | ✓ | ✓ | ✓ |
| Online evals on production | – | ✓ | ✓ |
| Guard runtime protection | – | 5M decisions | Unlimited |
| Automated red teaming | – | Nightly | Continuous |
| Human review routing | – | Your reviewers | + our specialists |
| Deployment | Managed | Managed | Managed, VPC, on-prem |
| SSO / SCIM / audit log | – | SSO | ✓ |
Pricing FAQ.
Is there a free tier?
Yes. The Developer plan is free for individuals and includes 25,000 graded cases per month, the full grader library and the CLI. No credit card.
How is Enterprise priced?
Enterprise is priced on evaluated volume, Guard traffic and deployment model (managed, single-tenant, VPC or on-prem). Most enterprise agreements are annual. Red-team engagements and custom data are scoped separately.
Do you offer pilots?
Yes. A 30-day paid pilot includes a threat model, a release-gating eval suite for one use case and a baseline report. The pilot fee is credited against an annual agreement.
Can we buy red teaming or data without the platform?
Yes. Engagements and datasets are available on their own. Findings are still delivered as a portable eval suite you can run with our free tier or your own tooling.
Find out what your model does before your users do.
A 45-minute briefing with our evaluation team. Bring a use case; leave with a threat model, a draft eval plan and a sense of where you stand.