From the model card
Fields below are copied from the tags and counters on the HuggingFace repository Qwen/Qwen2.5-Coder-7B-Instruct-AWQ 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-Coder-7B-Instruct
- quantized from Qwen/Qwen2.5-Coder-7B-Instruct
- Language tags
- English (en)
- Papers cited
- arXiv:2409.12186, arXiv:2309.00071, arXiv:2407.10671
- Downloads (HF counter at last fetch)
- 384,148
- Likes (HF counter at last fetch)
- 26
- Model card
- https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct-AWQ
Use cases
- Serving a capable code generation model on a single 8GB VRAM GPU
- Code completion, explanation, and debugging in production at lower cost
- Building coding assistants where the BF16 7B exceeds memory budget
- Batch code generation in CI pipelines with limited GPU resources
- Comparing AWQ quantisation impact on code model benchmark performance
Pros
- Official Alibaba AWQ release; quality benchmarks are reproducible
- Apache 2.0 license; TGI and Azure compatible
- AWQ typically preserves >98% of BF16 HumanEval accuracy
- 7B scale with AWQ fits single consumer-grade GPU
Cons
- Cannot be further fine-tuned post-AWQ quantisation
- AWQ dequantisation overhead increases token latency vs BF16 on fast hardware
- 7B context window is smaller than the 7B BF16 variant in some configurations
- Qwen2.5-Coder is superseded by Qwen3-Coder for new deployments
Tags
transformerssafetensorsqwen2text-generationcodecodeqwenchatqwenqwen-coderconversationalenarxiv:2409.12186arxiv:2309.00071arxiv:2407.10671base_model:Qwen/Qwen2.5-Coder-7B-Instructbase_model:quantized:Qwen/Qwen2.5-Coder-7B-Instructlicense:apache-2.0text-generation-inferenceendpoints_compatible4-bit