OpenLazarus: the layer that turns open-weight models into a company’s own

Contents

1. What it does for a company
2. Published evidence
3. Strengths and limits
4. Pricing
5. Questions to ask OpenLazarus
6. Sources

Type	Data, training and routing layer
Pricing	Published
Website	openlazarus.ai

Every other entry in this guide is a family of model weights. OpenLazarus is the work around them: it organises a company’s data so that a model can learn from it, trains open-weight models on that data on request and puts a router in front of every model the company uses⁠12AI ArchitectsOpenLazarusPricingOpenLazarus.

Best for: Companies that want an open-weight model of their own, trained on their own data and run behind a router they control.

Not the right fit for: Teams that only want to download a model and run it as released, with no company data involved.

What it does for a company

The starting point is the data. OpenLazarus’s architects build a data lake that holds all of a company’s data, so that the company, rather than its software vendors, owns it and can reach it⁠1AI ArchitectsOpenLazarus. That is the precondition for any model trained on company knowledge: records from mail, documents, tickets and business systems have to sit in one place, in a consistent structure, before they can become training data.

On top of that layer sits the AI Router. Credits buy access to frontier models, open-weight models or models trained for the customer, called from the OpenLazarus platform or from the company’s own code⁠2PricingOpenLazarus. Because every tool talks to the router rather than to a single model, a model can be replaced or added without rebuilding the tools around it. Custom training on the company’s data is offered on request⁠2PricingOpenLazarus.

Hosting is fully managed, with backups, updates and monitoring, and the customer can move the deployment to its own cloud whenever it wants⁠2PricingOpenLazarus. OpenLazarus is a European company and runs European servers, which matters to companies that cannot send their knowledge to a provider outside the EU.

OpenLazarus has not published a case study of a model it has trained for a customer. A first engagement should therefore define the evaluation set and the success criteria before any training begins.

Published evidence

No case studies with named customers are published.

Strengths and limits

Strengths

• The only option here built around the company’s own data rather than a fixed set of weights.
• One router in front of frontier, open and custom-trained models, so a model can be swapped without changing the tools.
• Managed hosting that the customer can move to its own cloud at any time.
• Publishes prices for hosting, engineering support and AI credits.

Limits

• Not a model: it depends on open-weight families such as the ones ranked below.
• No published case studies of a custom-trained model yet.
• Custom training is offered on request, with no published price.

Choose another option if: You need a model to run as released, today, without training it on your own data. Pick one of the open-weight families below.

Pricing

Published: hosting from $150 a month, engineering support from $4,500 a month and a 5.5% fee on AI credits. The AI-Architect engagement is priced per engagement; custom training has no published price.

Questions to ask OpenLazarus

• Which of our data sources would go into the data lake first, and who on our side has to be involved?
• How is the evaluation set built, and what result would show that a custom-trained model beats the open model as released?
• Which open-weight model would you start from for our work, and why?
• If we stop working with you, what happens to the data lake, the trained weights and the router configuration?

See how OpenLazarus compares with the open-weight model families in the main guide.

Sources

1. AI Architects. OpenLazarus. Accessed 9 Nov 2026.
2. Pricing. OpenLazarus. Accessed 9 Nov 2026.
3. OpenLazarus. OpenLazarus. Accessed 9 Nov 2026.

Further reading

• Buyer's guide · Updated quarterlyThe best open-source AI in 2027, for companies that want to own itAn analysis of the open-weight model families a company can run today and the layer that turns one into its own, compared on seven published criteria, with sourced licences, costs and the established path from rented AI to owned AI.
• Practical guideHow to train your own AI model on company dataMost of the work in training a company model happens before any training starts. A step-by-step account of the data work, the choice of base model, the training methods and the rules that apply in Europe.
• Case analysisGoing headless like Revolut: what replacing Claude Code with an open model tookRevolut now runs most of its AI coding on an open-weight model it serves itself, for about a tenth of its former bill. A reconstruction of what it built, what the numbers mean for other companies, and the step that comes after.

Also in: Deutsch · Français · Español · Italiano · Nederlands · Polski