Qwen3
Alibaba's open-weight Qwen3 language models, dense and mixture-of-experts, from 0.6B to 235B-A22B.
Apache 2.0 weights you download and serve yourself, with 256K-token context in the 2507 releases.
This repository documents Qwen3, the open-weight language model family from the Qwen team. It suits developers and researchers who want to run capable models on their own hardware.
What it is
The README lists dense models at 0.6B, 1.7B, 4B, 8B, 14B and 32B plus mixture-of-experts 30B-A3B and 235B-A22B. The 2507 updates add Instruct and Thinking variants with 256K-token context, extendable to 1 million tokens. Docs cover llama.cpp, Ollama, LM Studio, vLLM, SGLang and TGI.
Who it's for
It suits teams that can supply GPUs or a capable machine and want no per-token bill. It does not suit anyone expecting a one-click app: you download weights from Hugging Face or ModelScope and pick a serving stack.
Key features
Dense and MoE sizes
Models from 0.6B up to 235B-A22B.
2507 Instruct and Thinking
Updated variants in 235B-A22B, 30B-A3B and 4B.
Long context
256K-token context, extendable to 1 million tokens.
Local run guides
Docs for llama.cpp, Ollama and LM Studio.
Deployment guides
Docs for SGLang, vLLM and TGI.
Quantization docs
GPTQ, AWQ and GGUF guidance.
Up and running in 3 steps
Our honest verdict
What's great
- Weights are under Apache 2.0, which the README states for all open-weight models.
- Small sizes run on modest hardware, big sizes scale up.
- Works with the common local runtimes.
- Long-context variants are available.
Where it falls short
- The repository has no LICENSE file at its root; licences are in each Hugging Face repo.
- Larger models need serious GPU memory.
- You manage serving, updates and safety filtering yourself.
Pricing & license
The weights are free under Apache 2.0. Alibaba also sells hosted Qwen access separately; the repo shows no price, so your cost is the hardware you run it on.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Can I use Qwen3 commercially?
The README says all open-weight models are licensed under Apache 2.0; check the licence file in each Hugging Face repo.
Which sizes exist?
0.6B, 1.7B, 4B, 8B, 14B, 32B, 30B-A3B and 235B-A22B.
Can I run it locally?
Yes, the docs cover llama.cpp, Ollama and LM Studio.
