Fine-tune with the API or CLI

Learn how to tune Qwen models in Alibaba Cloud Model Studio using the API (HTTP) and the command line (shell). Model tuning involves three methods: supervised fine-tuning (SFT), continual pre-training (CPT), and direct preference optimization (DPO).

Prerequisites

• Review Introduction to model fine-tuning to understand the basic concepts, process, and data format requirements of model tuning.
• Activate the service and obtain an API key. For instructions, see Get an API key.
• Grant the Alibaba Cloud sub-account (RAM user) the necessary invocation, training, and deployment permissions.

Tuning file upload

Preparing fine-tuning files

For the data format specifications, packaging rules, size and quantity limits, and API upload quotas of each training method (SFT/DPO/CPT), see Tuning data upload rules. Overview of the file formats supported by each training method:

Upload fine-tuning files to Model Studio

curl --request POST \
'https://dashscope.aliyuncs.com/api/v1/files' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--form 'files=@"/path/to/your/file.jsonl"' \
--form 'purpose="fine-tune"' \
--form 'descriptions="a sample fine-tune data file for qwen"'

{
  "request_id":"xx",
  "data":{
    "uploaded_files":[{
      "file_id":"976bd01a-f30b-4414-86fd-50c54486e3ef",
      "name":"qwen-fine-tune-sample.jsonl"}],
  　"failed_uploads":[]}
 }

Model fine-tuning

Creating, querying, logging, snapshot management, and list/abort/delete operations for tuning jobs are all performed via API. Each interface's complete request/response parameters and examples are on the corresponding API page; this section only outlines the workflow.

Create a tuning job

Call the Create a tuning job interface (POST /api/v1/fine-tunes) to start training. For the request body (model / training_datasets / hyper_parameters / training_type, etc.) and response examples, see that API page.

Retrieve tuning job details

Use the job_id to query the task status. For details, see Query tuning task status.

Get tuning job logs

Training logs are retrieved via the Query tuning task logs interface (GET /api/v1/fine-tunes/{job_id}/logs).

Query and publish model parameter snapshots

Checkpoints produced by tuning can be listed and exported (published):

• List snapshots: see List Checkpoint (GET /api/v1/fine-tunes/{job_id}/checkpoints).
• Export (publish) a specific snapshot: see Export Checkpoint (GET /api/v1/fine-tunes/{job_id}/export/{checkpoint}?model_name=...), executed asynchronously; poll the status via the list interface.

More tuning operations

• List jobs: Query tuning job list
• Abort job: Cancel tuning job
• Delete job: Delete tuning job

API reference

The DashScope command-line invocation reference is included in this topic. For detailed API calls, see API details.

Model deployment and invocation

Model deployment

For more information about model deployment methods, see Deploy a model with the API.

curl "https://dashscope.aliyuncs.com/api/v1/deployments" \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
    "model_name": "qwen3-8b-ft-202511132025-0260",
    "plan": "lora",
    "capacity": 1,
    "name": "qwen3-8b-ft"
}'

curl "https://dashscope.aliyuncs.com/api/v1/deployments" \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
    "name": "my_qwen_plus",
    "model_name": "qwen-plus-2025-12-01",
    "plan": "mu",
    "deploy_spec": "MU1",
    "enable_thinking": true,
    "capacity": 4,
    "max_context_length": 10000,
    "rpm_limit": 500,
    "tpm_limit": 1000
}'

Query deployment status

When the deployment status is RUNNING, the model is currently available for invocation.

curl 'https://dashscope.aliyuncs.com/api/v1/deployments/<your_model_instance_id>' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY}  \
--header 'Content-Type: application/json'

For other deployment operations such as scaling and deprovisioning, see Model Deployment - API Details.

Model invocation

Once the model deployment status is RUNNING, you can use the fine-tuned model just like any other model.

You can also get the model code from the Dedicated Deployment console page.

For more usage methods and parameter settings, see the DashScope API Reference.

curl 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY}  \
--header 'Content-Type: application/json' \
--data '{
    "model": "<your_model_instance_id>",
    "input":{
        "messages":[
            {
                "role": "user",
                "content": "Who are you?"
            }
        ]
    },
    "parameters": {
        "result_format": "message"
    }
}'

Model evaluation

The model evaluation feature of Alibaba Cloud Model Studio is available only in the console. Go to the Model Evaluation page to evaluate model training results.

For more information, see Model Evaluation.