MODELS

Qwen/Qwen3-4B

by Qwen

modellead-sourcehugging-face-popular-modelssource:github.comtext-generationqwenqwen34bapache-2.0hugging-facealibaba-cloud

GitHub

Overview

Qwen3-4B is a 4B-parameter dense text-generation model in the Qwen3 family from the Qwen team at Alibaba Cloud.

Details

Qwen3-4B is one of the Qwen3 model sizes listed in the official Qwen3 repository, which describes Qwen3 as developed by the Qwen team at Alibaba Cloud and links to Hugging Face, ModelScope, the paper, blog, and documentation. The Qwen3 release blog states that Qwen3-4B is one of six dense Qwen3 models released as open-weighted models under the Apache 2.0 license, and notes architecture details including 36 layers and a 32K context length. The Hugging Face model page identifies Qwen/Qwen3-4B for text generation.

When to Use

Use for text-generation workflows where a Qwen3-family 4B dense model is appropriate. Use when you want an Apache-2.0-licensed, open-weight Qwen3 model size listed by the official Qwen3 sources. Evaluate it when comparing Qwen3 dense model sizes or testing Alibaba Cloud Qwen models through Hugging Face or the official repository links.

Getting Started

Open the Hugging Face model page at https://huggingface.co/Qwen/Qwen3-4B to inspect model files and usage metadata.
Review the official Qwen3 GitHub repository at https://github.com/QwenLM/Qwen3 for Qwen3 documentation links and model-family context.
Read the Qwen3 release blog at https://qwenlm.github.io/blog/qwen3/ for release notes, licensing, and architecture details.
Run a small text-generation evaluation before relying on the model in production.

Key Features

•4B dense model size in the Qwen3 family
•Text-generation model page on Hugging Face
•Open-weight release under the Apache 2.0 license according to the Qwen3 release blog
•32K context length and 36-layer architecture stated in the Qwen3 release blog
•Official Qwen3 repository links to Hugging Face, ModelScope, paper, blog, and documentation

Capabilities

•text-generation
•language-modeling
•open-weight-model

Last updated Jun 4, 2026