- Deploying fine-tuned models to serverless infrastructure
- Hosting the models on serverless infrastructure
- Deploying up to 100 fine-tuned models
- Usage costs on a per-token basis when the model is actually used
- The fine-tuning process itself, if applicable