// Custom models
Your weights. Our operations.
Upload weights or a container and FAR Labs optimizes, deploys, and scales dedicated inference in the target environment. Your data stays in your infrastructure and is not used to train third-party models.
Two ways to run a custom model
Both options run as dedicated deployments, either on FAR Labs-managed capacity or in an approved customer environment.
We adapt and serve
Bring an open-weight, fine-tuned, or proprietary model. FAR Labs prepares it for production inference, handles scaling and reliability, and lets you combine custom and catalogue models behind one API.
You keep fine-tuning
For models you continue to fine-tune, FAR Labs deploys updated weights or containers and keeps inference operations reliable. Serving expectations are reviewed as the model changes.
How deployment works
GPU-hour, metered by the minute
You pay the cost of GPUs reserved for you — an hourly rate for a specific GPU, metered by the minute — plus FAR Labs services.
Data center or your environment
Run in FAR Labs-managed capacity or in your approved environment. Specific placement, access, and data-boundary requirements are reviewed before deployment.
Batch at scale
Process large volumes of requests with an agreed completion deadline at a separate rate. Results remain available for later retrieval.
Bring your weights. Keep your data.
Tell us about your model and where it needs to run.