From the model card
Fields below are copied from the tags and counters on the HuggingFace repository Qwen/QwQ-32B at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- Qwen
- Pipeline tag
- text-generation
- Library
- Transformers
- Weight formats
- safetensors
- License tag
apache-2.0— read the license file in the repo before relying on it- Lineage
-
- base model Qwen/Qwen2.5-32B
- fine-tune of Qwen/Qwen2.5-32B
- Language tags
- English (en)
- Papers cited
- arXiv:2309.00071, arXiv:2412.15115
- Downloads (HF counter at last fetch)
- 380,357
- Likes (HF counter at last fetch)
- 2,957
- Model card
- https://huggingface.co/Qwen/QwQ-32B
Use cases
- Mathematical problem solving with explicit step-by-step reasoning
- Complex multi-step logical and scientific reasoning tasks
- Code generation that benefits from deliberate planning before writing
- Comparison against o1/o3-style reasoning models in open-weight settings
Pros
- 2,950 likes confirm it as a community flagship for open reasoning models
- Extensive chain-of-thought output enables auditing of reasoning steps
- Published arxiv references document benchmark performance
- text-generation-inference and conversational format support production serving
Cons
- Long CoT generation significantly increases latency and output token costs
- 32B requires A100 80 GB or 2×3090 for comfortable BF16 inference
- Reasoning length is not always calibrated — may over-explain trivial problems
- Qwen license terms for commercial use differ from Apache/MIT
Tags
transformerssafetensorsqwen2text-generationchatconversationalenarxiv:2309.00071arxiv:2412.15115base_model:Qwen/Qwen2.5-32Bbase_model:finetune:Qwen/Qwen2.5-32Blicense:apache-2.0eval-resultstext-generation-inferenceendpoints_compatibleregion:usdeploy:sagemakerdeploy:azure