LLMs come in two broad families. Proprietary models are accessed through a provider’s service. Open-weight or open-source models can be downloaded and run on infrastructure you control. Each has trade-offs.
Proprietary (hosted) models
- Pros: quick to start, no infrastructure to manage, strong general quality, regular improvements
- Cons: usage-based costs, data leaves your environment, limited control over changes
Open-source or open-weight models
- Pros: greater control, data can stay in your own environment, freedom to customise
- Cons: you need hardware and technical skill, you manage updates and security, and licence terms differ between models
Questions to ask
- Does regulation or a client contract require data to stay in a specific country or system?
- Do you have the technical team to host and monitor a model?
- Is your usage volume high and steady enough to justify your own infrastructure?
- Is the licence suitable for commercial use?
A hybrid approach
Many businesses use a hosted model for general tasks and a self-hosted one for sensitive data. Designing a flexible architecture lets you mix both.
Not sure which fits? Contact us or read about our AI & Automation services.