Qwen
Fine-tune Qwen3 on your data
Qwen's dense Qwen3 text models, in versions that answer directly and versions that reason before they answer. Tested on Tensorant on 7 November 2026: a run trained an adapter on Qwen3 4B Instruct 2507 and a deployment served it.
- Status
- Supported
- Input
- Text
- Kind
- Dense
- Trains with
- LoRA, QLoRA
- Suggested sizes
- LoRA from1 GPU, about 9 GB
Sizes
Sizes and GPUs
Each size with the GPUs one machine needs to train and to serve it. The figures are the estimates Tensorant checks a model against; a trial measures the real use. See GPU memory.
| Model | Parameters | Train with LoRA | Train with QLoRA | Serve | Status |
|---|---|---|---|---|---|
| Qwen/Qwen3-0.6B | 0.6B | 1 GPUabout 6 GB | 1 GPUabout 5 GB | 1, 2, 4 or 8 GPUsabout 4 GB in all | Supported |
| Qwen/Qwen3-1.7BReasons before answering | 1.7B | 1 GPUabout 9 GB | 1 GPUabout 5 GB | 1, 2, 4 or 8 GPUsabout 7 GB in all | Supported |
| Qwen/Qwen3-4B-Instruct-2507Small and quick to train | 4B | 1 GPUabout 14 GB | 1 GPUabout 7 GB | 1, 2, 4 or 8 GPUsabout 12 GB in all | Supported |
| Qwen/Qwen3-4BStronger reasoning | 4B | 1 GPUabout 14 GB | 1 GPUabout 7 GB | 1, 2, 4 or 8 GPUsabout 12 GB in all | Supported |
| Qwen/Qwen3-8B | 8B | 1 GPUabout 24 GB | 1 GPUabout 9 GB | 1, 2, 4 or 8 GPUsabout 22 GB in all | Supported |
| Qwen/Qwen3-14B | 14B | 1 GPUabout 40 GB | 1 GPUabout 14 GB | 1, 2, 4 or 8 GPUsabout 38 GB in all | Supported |
| Qwen/Qwen3-32B | 32B | 1 GPUabout 83 GB | 1 GPUabout 25 GB | 1, 2, 4 or 8 GPUsabout 81 GB in all | Supported |
- Training needs the memory shown on each GPU: more GPUs on a machine train faster, but each holds the whole model.
- Serve: the numbers of GPUs of one machine the base model can be split across, and the memory it needs in all.
Training
How it trains
Train Qwen3 with LoRA or QLoRA, by supervised fine-tuning (SFT), DPO on preference pairs or reinforcement learning (RL).
- RL trains with LoRA, on one GPU.
- A Qwen3 model that reasons writes its reasoning as part of its answer. In RL, the reasoning counts toward the longest sampled answer and is scored with the answer, so raise that limit.
Every training run, trials included, pays a training-token fee for the tokens it trains on, by the model's size.
Serving
How it serves
Publish a trained model under a name and call it through the OpenAI-compatible API, with the OpenAI SDK or any HTTP client. The name is what your application sends as model.
- A LoRA adapter you train serves on top of its base model, on the GPUs in the Serve column.
- The base model serves on its own too, in a test session or for a baseline. A deployment splits it across the GPUs of one machine: the Serve column lists the numbers each size allows.
Steps
Fine-tune Qwen3 on Tensorant
Create a project
Create a project with
Qwen/Qwen3-4B-Instruct-2507as its base model, or pick another size from the list. The model check reads the model before anything starts.Import your data
Upload files, paste text or import a Hugging Face dataset as sources.
Review and freeze a version
Review the examples and approve the ones to train on, then freeze them into a version with training, validation and test splits.
Score the base model
Evaluate the base model on the version's test split. This baseline is what a training run is compared with, and a full run needs one.
Run a trial, then train
A trial trains a small sample to check the settings, the GPU memory and the speed. Then start the full run. LoRA on
Qwen/Qwen3-4B-Instruct-2507needs 1 GPU, about 14 GB.Compare and publish
Evaluate the trained model on the same test split and compare it with the baseline. Then publish it and call it through the OpenAI-compatible API.
Fine-tuning
Fine-tune Qwen on RunPod with Tensorant
Build a support-ticket classifier with Qwen, review labeled examples, compare a LoRA adapter against the base model, and publish an API.
Docs
Supported models
Every family's architecture, status and notes, and what a base model needs.
Model family
Qwen3 MoE
Qwen's Qwen3 mixture-of-experts text models of 30B parameters, 3B of them active for each token, for chat, reasoning and code.
Model family
Llama
Meta's Llama 3.1 and 3.2 instruct text models, from ungated copies that need no access request.
Start with one dataset.
Create an account and import your data, or book a demo and we will show you the product.