#4848th of 52 in Open models
qwen3
Open-weight Qwen3 LLM family with dense and MoE models, 0.6B to 235B
- Stars
- 27.7k
- License
- no license
- Last commit
- Jan 2026
- Language
- Python
26,589 on 4 Apr 202627,698 on 31 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Qwen3 is Alibaba's open-weight LLM series, with dense models from 0.6B to 32B and MoE models at 30B-A3B and 235B-A22B. The repo is mainly a README that links to checkpoints on Hugging Face and ModelScope, plus run and deploy instructions for Transformers, llama.cpp, Ollama, vLLM, SGLang and TensorRT-LLM. The 2507 update splits the line into Instruct and Thinking variants with 256K context, extendable to 1M tokens.
Who it is for: Engineers picking open-weight LLMs to run locally or serve
Strengths
- Wide size range: 0.6B, 1.7B, 4B, 8B, 14B, 32B, 30B-A3B, 235B-A22B
- Run guides for llama.cpp, Ollama, LM Studio, MLX, OpenVINO, ExecuTorch and MNN
- Serving examples for vLLM, SGLang and TensorRT-LLM expose an OpenAI-compatible API
- 256K native context in 2507 models, extendable to 1M tokens
Weaknesses
- No license detected in the repo; terms must be checked on each model card
- Repo holds docs and snippets, not the training code or data
- Multi-step tool use with thinking models may degrade in vLLM and SGLang
- Qwen3-2507 evaluation blog is listed as coming soon
What it needs
- GPU optional
- Needs transformers>=4.51.0, llama.cpp>=b5401, vllm>=0.9.0, sglang>=0.4.6.post1
- Models: Transformers, GGUF, GPTQ, AWQ, MLX
Quick answers
Is qwen3 open source?
Its code is public on GitHub, but the repository has no license, so by default nobody may reuse it.
Does qwen3 need a GPU?
No. It runs on a CPU and can use a GPU when you have one.
Also in Open models
See all 52| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 95 out of 100 |
| 2 | llmfitRust tool that checks which LLMs fit your RAM, CPU and GPU | 87 out of 100 |
| 3 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 86 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 77 out of 100 |
| #5 | yolov5PyTorch YOLOv5 object detection, segmentation and classification models with training code | 75 out of 100 |