Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

Knowledge

What is a reasoning model?

A reasoning model thinks first and answers after, and that costs more time and tokens.

A person studies branching blocks on a laptop beside an hourglass.
A reasoning model uses extra reasoning before answering. Illustration: Denkbaar.

A reasoning model is a language model that thinks first and then answers. That costs more time and tokens.

What you read about it

  • On 28 November 2024 the Qwen team released QwQ-32B-Preview, an open model aimed at better reasoning. The team names its own limits: it can switch language and keep reasoning in circles: QwQ.
  • On 25 March 2025 Google described Gemini 2.5 Pro as a model that reasons before it answers: Gemini 2.5 Pro.
  • With OpenAI’s o3-mini (31 January 2025), developers could choose how much thinking the model spent on an answer: o3-mini.
A screen shows a loop between purple blocks.
A model can keep reasoning in circles. Illustration: Denkbaar.

Why it matters to you

Use a reasoning model where a mistake is costly, and not for every simple question. Write your own answer sheet first and compare settings on the same tasks. Also note how long the answer takes. A more expensive or slower setting need not be the default: let the task decide the quality you need.

Two people compare answer sheets beside two hourglasses.
Compare quality and waiting time using the same tasks. Illustration: Denkbaar.

Written on 13 November 2026, based on the articles and sources below.

Sources

  1. Qwen: QwQ: Reflect Deeply on the Boundaries of the Unknown
  2. Gemini 2.5 Pro
  3. OpenAI o3-mini

Topics: modelsreasoning