From the model card
Fields below are copied from the tags and counters on the HuggingFace repository cyankiwi/Qwen3.5-4B-AWQ-4bit 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)
- cyankiwi
- Pipeline tag
- image-text-to-text
- 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/Qwen3.5-4B
- quantized from Qwen/Qwen3.5-4B
- Downloads (HF counter at last fetch)
- 713,704
- Likes (HF counter at last fetch)
- 21
- Model card
- https://huggingface.co/cyankiwi/Qwen3.5-4B-AWQ-4bit
Use cases
- Multimodal inference on consumer GPUs with limited VRAM
- Visual question answering in local assistant applications
- Rapid prototyping of image+text pipelines without cloud APIs
- vLLM-based self-hosted multimodal serving
Pros
- ~4 GB VRAM footprint at AWQ 4-bit on a 4B model
- Multimodal image-text support at a very small model size
- Apache-2.0 license
- compressed-tensors format for vLLM integration
Cons
- 4B parameters constrains multimodal reasoning quality
- Community quantization — accuracy delta vs bf16 not independently benchmarked
- Vision quality degrades more with quantization than pure text tasks
- No GGUF variant available from this author
Tags
transformerssafetensorsqwen3_5image-text-to-textconversationalbase_model:Qwen/Qwen3.5-4Bbase_model:quantized:Qwen/Qwen3.5-4Blicense:apache-2.0endpoints_compatiblecompressed-tensorsregion:us