Customers can purchase a $5,000 annual license to run our models on infrastructure they control without the complexity of a custom enterprise agreement. To purchase our self-serve on-prem offering, log in to your Datalab account (sign up here) and visit the Subscription tab The license comes with:
  • Docker image for easy-to-run, single-GPU deployments.
  • Faster inference so you can achieve up to 20 pages/sec on an H100.
  • 12 months of updates with new functionality, bug fixes, and security updates.
  • Priority email support during business hours.
  • Commercial license for use in your product or services (without sublicensing)
  • One running instance, with up to two additional instances for $1,500 each.
  • 50,000 pages per month, with higher volumes available.
  • Self-serve checkout via Stripe for instance access.
When you purchase a license, you will get:
  • A license key: required to run the image.
  • A Google Cloud service account key: required to pull the image.
Running instances of the container must be able to reach https://www.datalab.to to start. They will send usage data at regular intervals, detailed here.

​
What’s the difference between Open Source and Datalab’s paid On-Prem options?

Our free open-source options (Marker and Surya) are ideal for research, personal use, and early-stage startups. The self-serve on-prem plan is a mid-tier option for lighter implementations that want a commercial license and simple deployment without heavy enterprise contracts. Enterprise is for teams that need scale, customization, and SLAs. Here’s a more detailed breakdown.
Free (Open Source)Self-Serve On-Prem ($5k/year)Enterprise (Custom)
Intended UseResearch, personal use, startups < $2M ARR/fundingLighter commercial workloads requiring privacyRegulated or high-scale deployments
LicenseGPL + custom RAILsCommercial license to use Marker & Surya on-prem (without sublicensing)Commercial license with custom rights as needed
DeploymentSelf-install from OSS reposStandalone Docker image for easy deployment and upgradesCustomer environment; Datalab-assisted rollout options
SupportCommunity onlyPriority email supportPremium support with SLAs
ContractingNoneClick-through agreement (no redlines)Custom agreements, security reviews
BillingFreeStripe credit card checkoutInvoice/PO, custom terms
Scale and PerformanceSelf-effortSingle-GPU support (20 pgs/sec on an H100) up to 50k pages/month includedThroughput/latency/accuracy tuning; custom page counts; custom rate limits
For higher page volume & GPU concurrency, fully-airgapped deployments, white glove support, and other custom needs, please fill out this form.

​
Try before you buy

We have two easy ways for customers to try our models: Our container image mimics the cloud-hosted API for a simple transition: