Fine-tune a model on your own data
Managed fine-tuning is in early access. You submit a training dataset, our team reviews the job, it trains with LoRA, and you are billed in Rupiah based on what the training actually cost.
How it works
From dataset to tuned model
Submit: upload a training dataset and choose a supported base model
Estimate: see the projected cost first; it is an estimate, never the charge
Review: our team approves every job before training starts
Train: LoRA fine-tuning on a managed GPU provider
Bill: charged in Rupiah from the training cost the provider reports
Why managed
Fine-tuning without the surprises
Reviewed before it runs
A person on our team approves each job, so a dataset mistake does not turn into a surprise bill.
LoRA training
Jobs train LoRA adapters, which are faster and cheaper than full fine-tuning.
Billed from real cost
The charge is based on the training cost the provider reports, not the up-front estimate. A job with no cost is not charged.
Paid balance only
Training is paid from your topped-up balance. The free signup credit cannot be used for fine-tuning.
Your data stays yours
We never opt in to letting the training provider use your data for its own models.
White-label
Results and bills come from Anoman, in Rupiah. No third-party provider name.
FAQ
Questions about fine-tuning
Is fine-tuning available now?
It is in early access. Request access and we will scope your first job with you.
How much does it cost?
It depends on the base model, dataset size and training settings. You see an estimate before the job runs, and you are billed in Rupiah from the actual training cost.
Can I call my tuned model through Anoman?
Not yet. Serving fine-tuned models through the Anoman API is on the roadmap.
Have a dataset ready?
Tell us about your use case and data, and we will set up your first fine-tuning job.