From the model card
Fields below are copied from the tags and counters on the HuggingFace repository Qwen/Qwen2.5-14B-Instruct 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/Qwen2.5-14B
- fine-tune of Qwen/Qwen2.5-14B
- Language tags
- English (en)
- Papers cited
- arXiv:2309.00071, arXiv:2407.10671
- Downloads (HF counter at last fetch)
- 2,685,103
- Likes (HF counter at last fetch)
- 363
- Model card
- https://huggingface.co/Qwen/Qwen2.5-14B-Instruct
Use cases
- Production-grade chat and instruction following at moderate cost
- Code generation and explanation tasks
- Multilingual document analysis with strong Chinese-English support
- Structured output extraction from complex documents
Pros
- Strong coding and math benchmarks for the 14B class
- Apache-2.0 licensed
- Long context support up to 128K tokens
- Broad deployment support across vLLM, SGLang, and Ollama
Cons
- BF16 14B requires ~28GB VRAM — typically needs two consumer GPUs or quantization
- Verbose by default without explicit length constraints
- Qwen2.5-32B provides noticeably better quality if hardware allows
- Limited native tool-call support compared to fine-tuned function-calling variants
Tags
transformerssafetensorsqwen2text-generationchatconversationalenarxiv:2309.00071arxiv:2407.10671base_model:Qwen/Qwen2.5-14Bbase_model:finetune:Qwen/Qwen2.5-14Blicense:apache-2.0text-generation-inferenceendpoints_compatibleregion:usdeploy:sagemakerdeploy:azure