Skip to content
LLMTR Docs

Qwen Plus Thinking Control

qwen/qwen-plus and qwen/qwen-plus-2025-07-14 are models with an optional thinking mode.

Thinking is off by default; turn it on with reasoning: true or :think. A plain request spends no thinking tokens and is billed at the non-thinking output price shown on the card. With thinking on, output is billed at the higher thinking price.

Request Result
qwen/qwen-plus (plain request) Does not think.
qwen/qwen-plus:think Thinks.
"reasoning": true Thinks.
"reasoning": { "enabled": true } Thinks.
"enable_thinking": true Thinks.
qwen/qwen-plus:fast, "reasoning": false, "reasoning": { "enabled": false }, "enable_thinking": false Does not think.
Terminal window
curl "$LLMTR_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer llmtr-your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen-plus:think",
    "messages": [{ "role": "user", "content": "What is 17 x 23? Work it out step by step." }],
    "max_tokens": 1024
  }'

With thinking on, thinking tokens do not count toward max_tokens; it limits only the answer. To limit the length of the reasoning chain, add thinking_budget to the request.

These models used to think by default. If an integration relies on that, add :think or reasoning: true to its requests. If you run into trouble, write to info@knowhy.co.