How to choose the right service
If you are unsure which service fits, start from the problem rather than the technique. Ask three questions. First, is there a specific, repeatable task you want the model to do? Second, do you have examples of that task done well, or the documents the model needs to understand? Third, where does the model need to run, and who is allowed to see the data?
If you have a clear task and you are currently paying for a large hosted model to do it, distillation is often the most direct way to cut costs. If general models simply do not perform well on your material, custom training is usually the better starting point. If you already have a model and need it in production, deployment and support is where to begin. And if you are an individual, the private client service keeps things simple.
Many projects combine services. A typical business engagement might start with a short discovery phase, move into data preparation and fine-tuning, and finish with deployment and a support plan. You can see indicative costs on our pricing page and the full engagement flow on our process page.
Whatever you choose, every engagement follows the same principles: you own your data and resulting weights, we never use your data for other clients, and we will tell you plainly if a simpler solution would serve you better. Read more about these commitments on our Responsible AI page.