Hugging Face - Qwen/QwQ-32B

As open-source reasoning models go, QwQ-32B stands out for focusing on hard problems rather than casual chat.

QwQ-32B is a large language model from Alibaba's Qwen team, built on Qwen2.5, that you download from Hugging Face and run yourself.

Within the Qwen series, QwQ is the branch for reasoning.

Compared with instruction-tuned models, QwQ does better on downstream tasks, especially difficult ones, so it fits maths, logic and multi-step questions more than quick small talk.

As a 32 billion parameter model, QwQ-32B is smaller than many reasoning models people compare it with, though you still need hardware that can handle that size.

There is one practical catch.

QwQ-32B requires the latest version of the Hugging Face transformers library, so an older setup has to be updated before the model will load.

If you already work with transformers, that is a small step.

If you want a chatbot that works out of the box, this is not it.