AI model selection and licence review
The right model per use case: performance, languages, licence and origin.
Overview
Lindstead tests candidate models on the organisation's own tasks and languages, then checks each licence and origin. The result is a shortlist auditors and the board can accept.
- ResultA tested model shortlist with licence and origin review.
- InputRepresentative tasks, languages, acceptance criteria and approved sample data.
Key questions
- Which open-weight models handle our tasks and languages well enough?
- May a company our size use this model commercially?
- Does the model's origin create policy or supply-chain risk?
- How far behind the frontier is the best model we can run ourselves?
Approach
- Use-case evaluation Candidate models run real tasks in the organisation's languages, scored per task.
- Licence review Commercial use, revenue or user thresholds, regional limits and redistribution. Some well-known licences exclude large companies or EU entities.
- Origin and supply-chain review Developer, jurisdiction, training disclosures and known tests of model behaviour, so origin becomes a policy decision.
Deliverables
- Model shortlist Models per use case, with test results and trade-offs.
- Licence review What each licence allows, in plain language.
- Origin assessment Supply-chain and policy points per candidate model.
Frequently asked questions
-
Open-weight means the trained weights are published, so the model can run in-house. The licence can still limit commercial use, company size or region, and training data is often undisclosed.
-
It depends on each model's licence and version. Lindstead's monthly Model Index lists licence terms for large companies per model.
-
Local weights stop data flowing to the developer, not behaviour trained into the model. Lindstead treats origin as a board decision and documents the trade-off.
Related
Discuss model selection with Lindstead.
A 30-minute meeting covers the use case, the requirements and a first plan.