OpenAI compatibility

​

Quick migration

import os
from openai import OpenAI

client = OpenAI(
  api_key=os.getenv("DASHSCOPE_API_KEY"),
  base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
  model="qwen3.6-plus",
  messages=[{"role": "user", "content": "Hello!"}],
)
print(completion.choices[0].message.content)

import OpenAI from "openai";

const openai = new OpenAI({
  apiKey: process.env.DASHSCOPE_API_KEY,
  baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
});

async function main() {
  const completion = await openai.chat.completions.create({
    model: "qwen3.6-plus",
    messages: [{ role: "user", content: "Hello!" }],
  });
  console.log(completion.choices[0].message.content);
}

main();

curl https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.6-plus",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

​

Supported APIs

​

Chat Completions

​

Qwen-specific parameters

​

Behavioral differences

• response_format supports json_object only (no json_schema).
• tool_choice supports auto, none, and specific function object ({"type": "function", "function": {"name": "..."}}). The required value is not supported.
• tools supports function type only.
• parallel_tool_calls defaults to false (OpenAI defaults to true).
• n supports 1-4 and is limited to specific models (qwen-plus, qwen-plus-character).
• web_search_options is not supported. Use extra_body.enable_search and extra_body.search_options instead.

​

Unsupported parameters

​

Responses API

• Built-in tools: web_search, code_interpreter, web_extractor, and image_search -- no external tool setup required.
• Simplified multi-turn: Pass previous_response_id instead of building a full message history.
• Conversation integration: Pair with the Conversations API for automatic context management.
• Session cache: Automatically caches context across turns to reduce latency and cost. Enable with the x-dashscope-session-cache: enable header. See Session cache.

​

Migrate from Chat Completions

1. Change base_url to https://dashscope-intl.aliyuncs.com/api/v2/apps/protocols/compatible-mode/v1 and the endpoint path from /v1/chat/completions to /v1/responses.
2. Read the response with output_text instead of choices[0].message.content.
3. For multi-turn conversations, pass previous_response_id instead of manually appending messages.

​

Embedding

• encoding_format: Only float is supported (default and only option).
• user: Not supported.
• dimensions: Available values depend on the model. For example, text-embedding-v4 supports 2,048, 1,536, 1,024 (default), 768, 512, 256, 128, and 64.

​

File API

• purpose must be file-extract (for document analysis with Qwen-Long/Qwen-Doc) or batch (for batch processing). OpenAI values like fine-tune and assistants are not supported.
• File content retrieval (GET /v1/files/{file_id}/content) is not supported.
• List filtering: The purpose and order parameters on GET /v1/files are not supported.
• Storage limits: 10,000 files, 100 GB total. Files never expire.

​

Batch API

• 50% cost discount compared to real-time pricing.
• completion_window: Supports 24h to 336h (14 days). Accepts "h" (hours) and "d" (days) units with integer values. OpenAI is fixed at 24h.
• Extra metadata: metadata.ds_name (task name) and metadata.ds_description (task description).
• Extra list filters: ds_name, input_file_ids, status, create_after, create_before.
• Input file limits: Up to 50,000 requests per file, 500 MB total, 6 MB per line. All requests in a file must use the same model.

​

Conversations API