Pricing

Priced per deployment. Not per GPU.

One serving endpoint, one price. Annual contracts, so the number procurement approves is the number on every invoice.

Evaluation

Free

For engineers proving it works

  • ✓Single GPU, non-production use
  • ✓Core compression and session resume
  • ✓Distributed as an executable
  • ✓Community support
  • ✓No signing or credential layer
Request the evaluation build

Team

£399

per deployment / month, billed annually

For a first production deployment

  • ✓One production deployment, up to 8 GPUs
  • ✓vLLM, SGLang, and NVIDIA NIM
  • ✓Full signing and W3C credential layer
  • ✓Encrypted persistence and resume without re-prefilling
  • ✓Standard support: email, 48-hour response
  • ✓Monthly usage and integrity reports
Talk to us

Enterprise

From £1,999

per month, custom annual contract

For banks, law firms, and regulated institutions

  • ✓Unlimited GPUs and deployments
  • ✓24/7 support with a named engineer
  • ✓Source code escrow with an independent agent
  • ✓Custom model validation and benchmarking
  • ✓Contractual indemnification, negotiated per agreement
  • ✓Quarterly business and integrity reviews
Contact sales

All production tiers are annual contracts, paid upfront. First-year maintenance included. Contracts available in GBP, EUR, or USD.

Professional services

Deployment is engineering work on your estate, and we price it as such.

Custom model integration

From £4,000 per model

Validation on your model architecture, benchmark runs on your corpus, integration testing on your serving stack. One-time fee.

Platform and GPU setup

From £2,000

Cloud GPU platforms or on-prem estates. One-time fee.

On-site deployment support

£1,500 per day plus travel

For air-gapped and restricted environments.

Annual maintenance

20% of annual license

Updates, security patches, and new serving-framework versions. Included in the first year of Team and Enterprise.

Questions buyers ask

What counts as a deployment?

One serving endpoint: a model instance (or replica group) behind a single endpoint, running Revyzor. You do not count GPUs, and scaling GPUs within a deployment does not change the price up to the tier limit.

Why per deployment and not per GPU?

Your platform team auto-scales GPUs; your bill should not auto-scale with them. Per-deployment pricing means procurement signs once and the number on the invoice is the number in the contract.

Is the source code available?

Revyzor ships as an executable. Enterprise contracts include source code escrow with an independent agent, released if Local Labs ceases trading, so your deployment is never stranded. Source access for security review is available under NDA as part of enterprise diligence.

Where does it run?

On your GPUs, inside your serving stack, under your keys. Revyzor is a plug-in to vLLM, SGLang, and NVIDIA NIM. Nothing is hosted by us and no data leaves your environment.

How does a pilot work?

Pilots run as a managed engagement on your model, your data, and your hardware. You see the results before committing to a tier. Pilot fees are credited against the first annual contract.

Start with a pilot

Your model, your data, your hardware. Results before commitment, and the pilot fee is credited against your first annual contract.

Request a Pilot