Qwen API access, OpenAI-compatible — and the enable_thinking switch

Qwen spans open-weight releases (Qwen2.5, Qwen3) and commercial tiers (qwen-max / plus / turbo). Instead of a DashScope key and Alibaba Cloud console, DaoXE puts Qwen behind the same key as GPT, Claude and DeepSeek. Call /v1/chat/completions with a qwen-* ID; see the cheapest APIs guide for where it fits a cost split.

Call Qwen through DaoXE#

Use /v1/chat/completions with an exact qwen-* ID from GET /v1/models. Thinking behavior changed across the family: Qwen3 hybrids toggle reasoning with the non-standard enable_thinking field, while newer Qwen3.5+ models think by default — send enable_thinking: false for plain output; some builds require streaming when thinking is on.

# Qwen "thinking": Qwen3 hybrids toggle it with enable_thinking (an extra
# field, not standard OpenAI); Qwen3.5+ models think BY DEFAULT — send
# enable_thinking: false for plain output. Some builds require streaming when
# thinking is ON.
curl --fail-with-body --show-error --silent \
  https://daoxe.com/v1/chat/completions \
  -H "Authorization: Bearer ${DAOXE_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_EXACT_MODEL_ID",
    "max_tokens": 64,
    "messages": [{"role": "user", "content": "Reply with OK."}],
    "enable_thinking": false
  }'

curl --fail-with-body --show-error --silent \
  https://daoxe.com/v1/chat/completions \
  -H "Authorization: Bearer ${DAOXE_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_EXACT_MODEL_ID",
    "max_tokens": 64,
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

Qwen-specific notes#

• enable_thinking is non-standard. It's an extra field beyond the OpenAI spec; not every client forwards it. Send it in the request body, or via your SDK's extra-body escape hatch.
• Newer Qwen3.5+ models think by default. On Qwen3 the enable_thinking switch decides; on Qwen3.5 thinking is on unless you explicitly send enable_thinking: false — don't assume a plain call returns plain output. (The old /think//no_think soft switches are gone in Qwen3.5.)
• Thinking may force streaming. With thinking ON, some Qwen builds only return via streaming — expect a stream, or turn thinking OFF for a simple call.
• Open vs commercial variants differ. qwen-max/plus/turbo and the open Qwen2.5/Qwen3 IDs behave and price differently — pick deliberately.
• Exact IDs only. Read qwen-* IDs from GET /v1/models; don't copy a version string from a blog.

What it costs#

Verify you actually get the model#

Prove the endpoint works before blaming the client — if this fails, no setting will fix it:

export DAOXE_API_KEY="your_api_key"

# List the exact model IDs your account can call
curl --fail-with-body --show-error --silent \
  https://daoxe.com/v1/models \
  -H "Authorization: Bearer ${DAOXE_API_KEY}"

Confirm connectivity, then diff a hard prompt against the official Qwen/DashScope API at temperature 0 to confirm the model class:

Frequently asked questions#