Skip to main content
Fine-tuning tailors a pretrained model to a smaller, targeted dataset so it performs better on a specific task or domain. Together AI handles the full lifecycle: data upload, training, hosting, and inference on a dedicated endpoint. Together AI currently supports two fine-tuning implementations:
  • LoRA: Trains a small set of adapter weights on top of the frozen base model. This is the default training mode, as it’s faster, cheaper, and the right choice for most use cases.
  • Full fine-tuning: Updates every weight in the base model. Uses more compute, but can outperform LoRA when the base behavior needs to shift substantially.
See LoRA vs. full fine-tuning to choose between them. This choice is separate from the training method you pick below.

​
Get started

Quickstart

Prepare your data, launch a LoRA job on Qwen3 8B, and evaluate the result.

Supported models

Browse every base model you can fine-tune, with context lengths and batch sizes.

Pricing

See how Together AI bills for training tokens and dedicated hosting.

Bring your own model

Fine-tune a model from Hugging Face that isn’t in the Together catalog.

​
Prepare your data

Your training file is JSONL or Parquet, formatted to match your task. See data preparation for the schemas, validation rules, and example datasets. 
from together import Together

client = Together()

train_file = client.files.upload(
    file="train.jsonl",
    purpose="fine-tune",
    check=True,
)
print(train_file.id)

​
Training methods

Supervised fine-tuning

Train on demonstration data with one target completion per example. The default method.

Preference fine-tuning

Align a model with rankings over preferred and dispreferred responses using DPO.

LoRA vs. full fine-tuning

Choose how much of the model to update, and tune LoRA’s rank and target modules.

​
Task-specific methods

Vision fine-tuning

Fine-tune vision-language models on image and text pairs.

Function calling

Train a model to invoke tools and structured functions reliably.

Reasoning fine-tuning

Train a reasoning model with chain-of-thought data.

​
Monitor and deploy

Monitor a job

Poll job status, then retrieve per-step loss and evaluation metrics.

Deploy your model

Serve your fine-tuned model on a dedicated endpoint or download it for local use.