Qwen models, local and online
Every guide we have written about the Qwen family, newest generation first. This page keeps its address when the next version arrives, so it is the one to bookmark.
Every guide
Qwen 3.8 Uncensored: Abliterated and Heretic Builds
Which abliterated, heretic and uncensored Qwen 3.8 27B builds exist, which GGUF file fits your VRAM, and how to run one locally. Sizes read off Hugging Face.
How to Run Qwen 3.8 27B Locally: VRAM, Quants and the Template Trap
Qwen 3.8 27B is out under Apache 2.0. Real GGUF sizes, the VRAM math for 12 GB and 24 GB cards, the broken chat template nobody warns you about, and vision setup.
Can You Run Qwen 3.8 Locally? The Honest Hardware Math
Qwen 3.8 Max has 2.4 trillion parameters. The honest math on running it locally, why the open weight Qwen 3.8 27B is the real story, and what to run today.
How to Run Qwen 3.8 Max: Every Option That Actually Works
Practical guide to using Qwen 3.8 Max today: the DashScope API with working code, OpenAI compatible endpoints, pricing, and the honest local answer.
Qwen 3.8 Max Explained: Alibaba's 2.4T Flagship Goes Open Weight
Qwen 3.8 Max is Alibaba's 2.4 trillion parameter MoE with 1M context and image plus video input. Specs, benchmarks, API pricing, and the open weights promise.
Qwen 3.8 Max vs Qwen 3.8 A95B, explained
Two Qwen 3.8 entries on every plan, almost the same price. The thinking switch is the whole difference, and on one of them it does not exist.
Qwen 3.8 vs Qwen 3.6: What Changed, and Which to Use
Qwen 3.8 Max vs Qwen 3.6: 2.4T vs 27B and 35B, 1M vs 256K context, the benchmark jumps, and which Qwen you can actually use today, hosted or locally.
How to run Qwen 3.8 without a GPU
Both big Qwen 3.8 checkpoints run on every plan and on a credit pack, no subscription. Credits per token, and the 27B from the same wave you can run at home.
How to Run Qwen 3.6 Locally: 27B Dense, 35B MoE, and Coding Variants
Two ways to run Qwen 3.6 on your own GPU: Ollama or a one-click installer. Direct GGUF downloads and the right quant for 8, 12, 16 or 24 GB VRAM, free.
Qwen-Image 2.1, uncensored on your own card
Qwen-Image 2.1 on your own card: the one click bundle in Locally Uncensored, the uncensored community builds, GGUF file sizes, LoRAs and the research license.
How to Run Qwen-Image 2.1 Locally: One Model That Draws and Edits
Run Qwen-Image 2.1 on your own graphics card with Locally Uncensored 3.0.1: the three files and their sizes, the ComfyUI version it needs, the Image lane, the maskless Edit lane, and the research license that rules out commercial work.
Qwen-Image 2.1 Explained: 7B, Alpha Channel, Ten References, and a New License
What changed in Qwen-Image 2.1: a 7B image model where the first one had 20B, a VAE with an alpha channel, ten reference images in a pass, one model instead of two, and a research license where the series used to be Apache 2.0.
FAQ
Which Qwen model can I run at home?
Qwen 3.8 27B is the one that fits a card you can buy. A 4 bit file is about 17 GB, so a 24 GB card is the comfortable home. Qwen 3.8 Max has 2.4 trillion parameters and does not fit consumer hardware.
Is there an uncensored Qwen?
Yes. Abliterated and heretic builds of Qwen 3.8 27B have been on Hugging Face since August 14, 2026. The Qwen 3.8 uncensored page lists them with file sizes.
Can I use Qwen without a GPU?
Yes, through a hosted service. Qwen 3.8 27B, Qwen 3.8 A95B and Qwen 3.8 Max are in the LU Cloud catalog, which opens in a browser.