Skip to content
Part of Dynamis Group
Client Portal

Lease the model, or own it outright.

Two commercial paths for a delivered model: hosted under a monthly fee, or a single build-and-deliver fee with the weights, training configuration and technical report transferred to you on completion.

Dynamis Labs — Licensing offers two commercial paths for delivered computer-vision models. Lease — monthly fee, hosted on the client's cloud, quarterly re-training and SLA-backed inference included. Own outright — single build-and-deliver fee with weights, training configuration, evaluation harness and technical report transferred on completion. Lease is convertible to ownership at any time. Both paths cover commercial use for the licensed enterprise.

  1. Research
  2. Prototyping
  3. Production
  4. Licensing

Commercial options

Bespoke models for specific enterprise problems.

Off-the-shelf vision APIs solve common tasks on common data. They rarely solve a regulator’s integrity brief, a broadcaster’s provenance check, or an insurer’s adjudication workflow. We build models to fit the problem — and structure the engagement so the client can own them.

Option A

Lease

Model-as-a-service lease

The trained model is hosted and maintained by Dynamis Labs on your chosen cloud. You pay a predictable monthly fee; we handle re-training cadence, evaluation and drift monitoring.

  • Predictable monthly fee
  • Quarterly re-training included
  • SLA-backed inference endpoint
  • Transferable to ownership at any time

Option B

Own outright

Outright model ownership

A single build-and-deliver fee. You receive the trained weights, evaluation harness, training configuration and technical report. The model is yours — operated on your infrastructure, on your terms.

  • Weights & code transfer on delivery
  • No recurring licensing
  • Technical report & training runbook
  • Optional research-retainer support

Both options cover commercial use for the licensed enterprise. Model re-distribution, model marketplaces and derivative model training on the delivered weights require a separate agreement.

Common questions

FAQs

Here are some of our most frequently asked questions. Can't find what you're looking for? Reach out to our support team.

What is the difference between leasing and owning the model?
Leasing — Dynamis Labs hosts and maintains the trained model on the cloud you choose, with a monthly fee covering inference, quarterly re-training and drift monitoring. Owning outright — single build-and-deliver fee, with weights, training configuration, evaluation harness and technical report transferred on completion. Leasing converts to ownership at any time on a defined formula.
What rights do I get with outright ownership?
Commercial use within the licensed enterprise — production deployment, internal tooling, customer-facing inference, derivative product development. What requires separate agreement: model redistribution, model marketplaces, derivative-model training that uses the delivered weights as initialisation, and any onward licensing to third parties. The base licence covers what most enterprises actually do with a model.
Can a third party validate the model?
Yes — the evaluation harness is delivered with the model under both lease and outright purchase, so an external validator (an internal audit team, an academic group, a regulator-engaged third party) can re-run the evaluation independently. Reproducibility of the evaluation is non-negotiable; reproducibility of the training is included with outright ownership.
Can I lease and convert to ownership later?
Yes. The lease agreement includes a transfer-to-ownership clause: at any time the client can pay the difference between accumulated lease fees and the outright purchase price (less a defined credit), and take delivery of weights, configuration and the technical report. No penalty, no notice period beyond that of the underlying lease.
What does the technical report include?
Architecture description (backbone, head, training objective), training configuration (hyperparameters, optimiser, schedule, data splits), evaluation methodology and held-out results, inference performance (latency, throughput, hardware), failure-mode register, and a written runbook for re-training. Written so an inheriting team can re-run the build from scratch.

Start a conversation

One architect, one inbox.

Bring us the situation. We’ll pair you with a solution architect and write back — no hand-offs across divisions, no sales cadence.

Get in touch