Thinking

A thinking model reasons before answering and returns the reasoning beside the answer.

Thinking on and off

Thinking is on by default. "chat_template_kwargs": {"enable_thinking": false} turns thinking off.

Use thinking for maths, code and multi-step questions. Turn thinking off for chat, extraction, classification, tool calls and many parallel requests.

chat_template_kwargs is not an OpenAI parameter. The Python SDK sends the field through extra_body; curl and fetch send the field at the top level of the body.

Example

The example asks the same question with thinking on and off.

curl https://api.albastov.com/v1/chat/completions \
  -H "Authorization: Bearer $ALBASTOV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.8-27b",
    "max_tokens": 400,
    "messages": [{"role": "user", "content": "Is 91 prime? Answer yes or no."}]
  }'

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.albastov.com/v1", api_key=os.environ["ALBASTOV_API_KEY"])
question = [{"role": "user", "content": "Is 91 prime? Answer yes or no."}]

on = client.chat.completions.create(model="qwen3.8-27b", messages=question, max_tokens=400)
msg = on.choices[0].message
print("reasoning:", getattr(msg, "reasoning", None))
print("answer:", msg.content.strip())
print("reasoning tokens:", on.usage.completion_tokens_details.reasoning_tokens)

off = client.chat.completions.create(
    model="qwen3.8-27b", messages=question, max_tokens=400,
    extra_body={"chat_template_kwargs": {"enable_thinking": False}},
)
print("answer without thinking:", off.choices[0].message.content.strip())

{
  "message": {
    "role": "assistant",
    "content": "\n\nNo",
    "reasoning": "We need answer user: \"Is 91 prime? Answer yes or no.\" Need final yes or no. 91=7*13 not prime. Answer no. Ensure only yes/no maybe \"No\".\n"
  },
  "usage": {
    "prompt_tokens": 63,
    "completion_tokens": 50,
    "total_tokens": 113,
    "completion_tokens_details": {"reasoning_tokens": 46}
  }
}

Token budget

max_tokens includes reasoning tokens. A low max_tokens ends the response during the reasoning, with finish_reason: "length" and empty content. Set max_tokens to 1,000 or more for hard questions, and stream long reasoning: a non-streamed request without a response after about 100 s fails with 524.

Reasoning field

The reasoning field is reasoning (delta.reasoning in a stream). Clients written for other servers may read reasoning_content instead. The OpenAI Python SDK returns reasoning as an attribute: read message.reasoning.

• content can start with blank lines after the reasoning. Trim content.
• Send earlier assistant turns with content only, without reasoning.

Model support

Every chat model reasons (capabilities.thinking is true in GET /v1/models). The enable_thinking switch is confirmed on qwen3.8-27b, qwen3.8-27b-fast and qwen3.8-27b-uncensored. qwen3.8-27b-uncensored reasons more briefly than qwen3.8-27b.

Reference

• Chat completions reference: chat_template_kwargs, max_tokens and the reasoning response fields.