Pricing

Free to adopt. Paid where it ships value.

The open-source SDK runs every environment locally. Upgrade when you need hosted calibration, team audit history, signed attestation receipts, or self-hosted deployment inside your VPC.

Free SDK
$0forever

For developer adoption and individual evaluations. Run every environment locally on the Apache-2.0 SDK. No account, no telemetry, no feature gates.

pip install
  • 25 evaluation environments across 6 capability domains
  • Objective ground truth (executable or closed-form) on every task
  • Contamination-resistant by construction
  • Apache-2.0 licence, source on GitHub
  • Community support via GitHub issues
Need hosted calibration? Pay-as-you-go via API tier ↓
Popular
Pro
$499/ month

For team workflows: hosted calibration, batch evaluations, audit history, and priority support. Standard SLA-backed inbox.

Start Pro
  • Everything in Free + API access
  • 1M evaluation traces / month included
  • Signed audit history (90 days retention)
  • 1,000 RPM, parallel batch jobs
  • Email support, 24h response SLA
  • 15% discount on annual billing
Enterprise
Customfrom $50K / year

For frontier AI labs, model vendors, agent infrastructure companies, and regulated AI teams. Self-hosted or VPC, founder-led integration, and the attestation work to sit inside your training, procurement, and release-gate workflows.

Book enterprise deployment
  • Everything in Pro, unlimited traces
  • Self-hosted or VPC deployment available
  • V-Certified Bronze, Silver, or Gold attestation
  • Signed X.509 attestation certificates (available for enterprise pilots)
  • Dedicated Slack channel, 4h response SLA
  • Founder-led integration into training, procurement, or release workflows
V-Certified programme

Three certification tiers designed for procurement teams.

Each tier ships a cryptographically-signed X.509 attestation certificate, available for enterprise pilots. Independently auditable when enabled.

Bronze$4,999 / yr

Annual self-attested calibration. CI integration, signed certificate, registry listing.

Silver$24,999 / yr

Quarterly third-party validation by Verifiable Labs. Custom env additions included.

Gold$99,999 / yr

Annual independent audit, public report, full reproducibility manifest, 1h response SLA.

Compare every tier

Every feature, every tier — side by side.

Free SDK$0foreverAPI$0.10/ 1K tracesPro$499/ monthEnterpriseCustomfrom $50K / yrSelf-hostedCustomfrom $250K / yr
Core
25 environments across 6 domains
Classical / executable ground truth
Apache-2.0 SDK + source
Hosted calibration API
Pay-as-you-go
1M /mo
Unlimited
Self-hosted
Conformal-prediction intervals
Limits & support
Trace retention
30 days
90 days
1 year
Forever
Rate limit (RPM)
100
1,000
10,000
Unlimited
Dedicated support channel
Community
Email
Email · 24h
Slack · 4h
Founders · 1h
V-Certified attestation
V-Certified attestation
Signed X.509 certificate
Public registry listing
Deployment
Air-gapped / VPC deployment
Source licence for hosted services
Pricing FAQ

Common questions before you pick a tier.

Anything missing? Email [email protected] — we'll add it here.

How fast can we get started?

Five minutes from `pip install` to your first calibrated score. No account setup, no demo call required — just install the Apache-2.0 SDK and run.

Do you offer a free trial?

The Apache-2.0 SDK is free forever. Run all 25 environments locally, see the calibrated rewards, no account required. Upgrade to Pro when you need hosted calibration, batch jobs, or attestation receipts.

Do you integrate with W&B, MLflow, or LangChain?

The SDK emits trace JSONL files compatible with any pipeline that reads JSON. Direct integrations with Weights & Biases and MLflow ship Q2 2026. Custom integrations are available on Enterprise.

Is the SDK really free forever?

Yes. Apache-2.0, no telemetry, no key validation. Self-host the entire stack if you want — the source for the hosted services is also available under Enterprise.

What counts as a trace?

One end-to-end episode through an environment, including the model rollout, classical baseline, reward computation, and conformal-interval calculation. A standard 30-episode audit consumes 30 traces.

Do you offer annual / academic discounts?

Pro: 15% off on annual billing. Academic / non-profit: 50% off Pro, free Bronze cert for one paper per year. Reach out to [email protected].

How does V-Certified differ from a regular audit?

A regular audit is a one-off PDF. V-Certified is a signed certificate registered in a public chain — third parties can verify the attestation without trusting either party.

Start with the SDK. Upgrade when procurement signs the receipt.

The open-source SDK gives every environment local. Add hosted calibration when you need bounded evidence on every reward, signed attestation when procurement asks.