LLM EXPLORER 66,845 MODELS INDEXED

Qwen3.5 4B QUASAR NVFP4 by QUASAR-QAT

By QUASAR-QAT · 190 downloads
HF Repo ↗ Copy URL

Qwen3.5 4B QUASAR NVFP4 is an open-source language model by QUASAR-QAT. Features: 4b LLM, VRAM: 4.2GB, License: apache-2.0, LLM Explorer Score: 0.26.

  Arxiv:2608.13966   4-bit precision   4bit   8-bit Base model:quantized:qwen/qwen...   Base model:qwen/qwen3.5-4b   Compressed-tensors   Conversational   Endpoints compatible   Eval-results   Fp4   Image-text-to-text   Long-context   Mtp   Multimodal   Nvfp4   Quantization   Quantization-aware-training   Quantized   Quasar   Qwen3.5   Qwen3.5-4b   Qwen3 5   Reasoning   Region:us   Safetensors   Speculative-decoding   Thinking   Vision   Vision-language   Vllm   W4a16
Model Card on HF 🤗: https://huggingface.co/QUASAR-QAT/Qwen3.5-4B-QUASAR-NVFP4 

Qwen3.5 4B QUASAR NVFP4 Parameters and Internals

LLM NameQwen3.5 4B QUASAR NVFP4
Repository 🤗https://huggingface.co/QUASAR-QAT/Qwen3.5-4B-QUASAR-NVFP4 
Base Model(s)  Qwen/Qwen3.5-4B   Qwen/Qwen3.5-4B
Model Size4b
Required VRAM4.2 GB
Updated2026-09-14
MaintainerQUASAR-QAT
Model Typeqwen3_5
Model Files  4.2 GB
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version4.57.0.dev0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Qwen3.5 4B QUASAR NVFP4

Best Alternatives
Context / RAM
Downloads
Likes
AgentHorse 4B256K / 8.4 GB9737
Qwen3.5 4B0K / 9.3 GB6575502801
NuExtract30K / 0.2 GB441570324
LightOnOCR 3 4B0K / 9.1 GB239571
Agents A1 4B0K / 9.1 GB17757872
...en3.5 4B 3D Coder MTP Repacked0K / 9.3 GB21328
FrogNano 4B 26090K / 9.3 GB733113
RuBio 4B0K / 9.3 GB38810
Qwen3.5 4B Base0K / 9.3 GB22781479
LightOnOCR 3 4B MTP Repacked0K / 9.3 GB6621
Note: green Score (e.g. "73.2") means that the model is better than QUASAR-QAT/Qwen3.5-4B-QUASAR-NVFP4.