From the model card
Fields below are copied from the tags and counters on the HuggingFace repository deepseek-ai/DeepSeek-R1-0528-Qwen3-8B 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)
- deepseek-ai
- Pipeline tag
- text-generation
- Library
- Transformers
- Weight formats
- safetensors
- License tag
mit— read the license file in the repo before relying on it- Papers cited
- arXiv:2501.12948
- Downloads (HF counter at last fetch)
- 976,417
- Likes (HF counter at last fetch)
- 1,083
- Model card
- https://huggingface.co/deepseek-ai/DeepSeek-R1-0528-Qwen3-8B
Use cases
- Multi-step mathematical problem solving with chain-of-thought
- Code generation tasks requiring intermediate reasoning steps
- Question answering over structured or semi-structured data
- Prototyping reasoning pipelines on consumer-grade hardware
- Academic research into distilled reasoning model behavior
Pros
- MIT license allows commercial and derivative use without restrictions
- 8B parameter count fits on a single 24GB GPU in fp16
- Inherits Qwen3 architecture improvements including extended context support
- Compatible with text-generation-inference for optimized serving
- Over 600K downloads signals broad community validation
Cons
- Distillation from a larger model can introduce capability gaps on edge-case reasoning
- No official fine-tuning dataset disclosed, limiting reproducibility of training
- Chain-of-thought verbosity increases token cost at inference time
- Smaller capacity than full R1 variants means weaker performance on hard benchmarks
- Community-reported evaluations may not reflect your domain-specific performance
Tags
transformerssafetensorsqwen3text-generationconversationalarxiv:2501.12948license:mittext-generation-inferenceendpoints_compatibleregion:us