What are the hardware requirements for Qwen3.5 4B MLX?

Qwen3.5 4B MLX requires approximately 3 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Qwen3.5 4B MLX and how large is it?

Qwen3.5 4B MLX is developed by nace-ai, a model with 4b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Qwen3.5 4B MLX?

Qwen3.5 4B MLX is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Qwen3.5 4B MLX by nace-ai — VRAM 3GB

Name: Qwen3.5 4B MLX
Author: nace-ai

Qwen3.5 4B MLX is an open-source language model by nace-ai. Features: 4b LLM, VRAM: 3GB, License: apache-2.0, LLM Explorer Score: 0.3.

4-bit Apple-silicon Base model:quantized:qwen/qwen... Base model:qwen/qwen3.5-4b Conversational Image-text-to-text Kapitsa Mlx Q4 Quantized Qwen3 5 Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/nace-ai/Qwen3.5-4B-MLX

Qwen3.5 4B MLX Parameters and Internals

LLM Name	Qwen3.5 4B MLX
Repository 🤗	https://huggingface.co/nace-ai/Qwen3.5-4B-MLX
Base Model(s)	Qwen/Qwen3.5-4B Qwen/Qwen3.5-4B
Model Size	4b
Required VRAM	3 GB
Updated	2026-11-13
Maintainer	nace-ai
Model Type	qwen3_5
Model Files	3.0 GB
Model Architecture	Qwen3_5ForConditionalGeneration
License	apache-2.0
Model Max Length	262144
Transformers Version	4.57.0.dev0
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Errors	replace

Rank the Qwen3.5 4B MLX capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Qwen3.5 4B MLX

Best Alternatives	Context / RAM	Downloads	Likes
AgentHorse 4B	256K / 8.4 GB	973	7
Qwen3.5 4B	0K / 9.3 GB	6575502	801
NuExtract3	0K / 0.2 GB	441570	324
Agents A1 4B	0K / 9.1 GB	177578	72
LightOnOCR 3 4B	0K / 9.1 GB	2395	71
SingVL S1 4B	0K / 9.1 GB	446	3
...en3.5 4B 3D Coder MTP Repacked	0K / 9.3 GB	2132	8
Rein NPC 4B	0K / 10.6 GB	1366	3
FrogNano 4B 2609	0K / 9.3 GB	733	113
...5 4B GLB 3D Coder MTP Repacked	0K / 17.7 GB	2557	1

Note: green Score (e.g. "73.2") means that the model is better than nace-ai/Qwen3.5-4B-MLX.