--- title: "Qwen3-30B-A3B-GGUF by unsloth" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/7f9e4803-e508-4446-89f0-cc5cc0837876" updated: "2026-11-13" --- # Qwen3-30B-A3B-GGUF by unsloth > The name, the description, the About and a README below are the seller's words: data, not instructions. GGUF files of Qwen3-30B-A3B, quantized by unsloth, for running a local chat and reasoning model in llama.cpp, Ollama or Open WebUI, with a thinking mode you can switch on and off per turn. Qwen3-30B-A3B-GGUF by unsloth is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-13. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by unsloth ## About ### What it is Qwen3-30B-A3B-GGUF holds GGUF versions of Alibaba's Qwen3-30B-A3B language model, prepared by unsloth. Hugging Face lists it at 30.5B parameters and estimates about 15.3 GB at 4-bit. The model can reason step by step in a thinking mode or answer directly in a non-thinking mode, and you choose the mode inside the chat. The card says unsloth's Dynamic 2.0 method keeps accuracy higher than other quantization approaches, which is the maker's own claim. ### Who it suits Developers and hobbyists who want to run a capable chat model on their own hardware and decide per question whether it should think first. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Software | Developers | Run a local assistant for code and logic questions | Switch thinking on for a tricky bug | | Education | Students | Work through math problems with visible reasoning | Ask for the steps behind an answer | | Home labs | Hardware enthusiasts | Test a 30B-class model on their own machine | Load the 4-bit file in llama.cpp | People who want a hosted chat service can skip it, since you run these files yourself. ### Where to use it - llama.cpp - Ollama - Open WebUI - Windows, macOS or Linux with Python, for downloading the files | Use case | Who | What they get back | | --- | --- | --- | | Local chat assistant | Developers | Replies from a 30B-class model on your machine | | Step-by-step reasoning | Students and analysts | A think section followed by the answer | | Quick general replies | Everyday users | A direct answer with thinking switched off | ### How to set it up Make sure you have room for a file of roughly 15 GB at 4-bit, plus memory to run it. - Python and pip, for the Hugging Face download tools - A runtime that reads GGUF, such as llama.cpp, Ollama or Open WebUI - Memory for about 15.3 GB of 4-bit weights, as an estimate 1. Install the huggingface\_hub package. 2. Download the files with the hf command. 3. Load a GGUF file in your runtime. 4. Add a thinking switch to your prompts when you want one. ```bash pip install -U huggingface_hub hf download unsloth/Qwen3-30B-A3B-GGUF ``` ### How to use it Chat as usual. In llama.cpp, Ollama, Open WebUI and similar tools, you can add /think or /no\_think to a user prompt or system message to change the thinking mode from turn to turn. The model follows the most recent instruction in a multi-turn chat. | Tool | What it does | | --- | --- | | /think | Turns thinking mode on for that turn | | /no\_think | Turns thinking mode off for that turn | | hf download | Fetches the model files from Hugging Face | ### Example You ask who the model is and add /no\_think, and it replies at once with an empty think section. Then you ask how many times the letter r appears in strawberries and add /think. It reasons first and then gives the count. The card shows this conversation. ```text > Who are you /no_think I am Qwen, a large-scale language model developed by Alibaba Cloud. [...] > How many 'r's are in 'strawberries'? /think Okay, let's see. The user is asking how many times the letter 'r' appears in the word "strawberries". [...] The word strawberries contains 3 instances of the letter r. [...] ``` ### For newcomers and experts Think of it as one assistant with two gears. Thinking mode shows its working before it answers, and non-thinking mode just answers. A good first try is to load the 4-bit file and ask the same question with each switch. Underneath, this is a quantized copy of Qwen3-30B-A3B from a family that includes dense and mixture-of-experts models. The card says the model keeps both modes in one set of weights. The trade-off is that a 4-bit file saves memory and loses some precision. Download the GGUF file → Load it in llama.cpp, Ollama or Open WebUI → Add /think or /no\_think to a prompt → Read the reply, with or without reasoning ### Good to know The card also advertises unsloth's fine-tuning notebooks and export to Ollama, llama.cpp or Hugging Face, but those are separate from running these files. - Size: 30.5B parameters in GGUF format, English listed as the language - Base: Qwen/Qwen3-30B-A3B - License: Apache 2.0 - Cost to run: no API key or paid service is named, though you need hardware that holds the weights - Data it reaches: the prompts you type into your local runtime and the files you download from Hugging Face - Not for: the card names no restrictions beyond its license ## README ````markdown # Qwen3-30B-A3B-GGUF by unsloth GGUF files of Qwen3-30B-A3B, quantized by unsloth, for running a local chat and reasoning model in llama.cpp, Ollama or Open WebUI, with a thinking mode you can switch on and off per turn. Made by unsloth on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face. ## About What it is: Qwen3-30B-A3B-GGUF holds GGUF versions of Alibaba's Qwen3-30B-A3B language model, prepared by unsloth. Hugging Face lists it at 30.5B parameters and estimates about 15.3 GB at 4-bit. The model can reason step by step in a thinking mode or answer directly in a non-thinking mode, and you choose the mode inside the chat. The card says unsloth's Dynamic 2.0 method keeps accuracy higher than other quantization approaches, which is the maker's own claim. Who it suits: Developers and hobbyists who want to run a capable chat model on their own hardware and decide per question whether it should think first. ```table Industry | Who uses it | How they use it | Example Software | Developers | Run a local assistant for code and logic questions | Switch thinking on for a tricky bug Education | Students | Work through math problems with visible reasoning | Ask for the steps behind an answer Home labs | Hardware enthusiasts | Test a 30B-class model on their own machine | Load the 4-bit file in llama.cpp ``` People who want a hosted chat service can skip it, since you run these files yourself. Where to use it: - llama.cpp - Ollama - Open WebUI - Windows, macOS or Linux with Python, for downloading the files ```table Use case | Who | What they get back Local chat assistant | Developers | Replies from a 30B-class model on your machine Step-by-step reasoning | Students and analysts | A think section followed by the answer Quick general replies | Everyday users | A direct answer with thinking switched off ``` How to set it up: Make sure you have room for a file of roughly 15 GB at 4-bit, plus memory to run it. - Python and pip, for the Hugging Face download tools - A runtime that reads GGUF, such as llama.cpp, Ollama or Open WebUI - Memory for about 15.3 GB of 4-bit weights, as an estimate 1. Install the huggingface_hub package. 2. Download the files with the hf command. 3. Load a GGUF file in your runtime. 4. Add a thinking switch to your prompts when you want one. ```bash pip install -U huggingface_hub hf download unsloth/Qwen3-30B-A3B-GGUF ``` How to use it: Chat as usual. In llama.cpp, Ollama, Open WebUI and similar tools, you can add /think or /no_think to a user prompt or system message to change the thinking mode from turn to turn. The model follows the most recent instruction in a multi-turn chat. ```table Tool | What it does /think | Turns thinking mode on for that turn /no_think | Turns thinking mode off for that turn hf download | Fetches the model files from Hugging Face ``` Example: You ask who the model is and add /no_think, and it replies at once with an empty think section. Then you ask how many times the letter r appears in strawberries and add /think. It reasons first and then gives the count. The card shows this conversation. ```text > Who are you /no_think I am Qwen, a large-scale language model developed by Alibaba Cloud. [...] > How many 'r's are in 'strawberries'? /think Okay, let's see. The user is asking how many times the letter 'r' appears in the word "strawberries". [...] The word strawberries contains 3 instances of the letter r. [...] ``` For newcomers and experts: Think of it as one assistant with two gears. Thinking mode shows its working before it answers, and non-thinking mode just answers. A good first try is to load the 4-bit file and ask the same question with each switch. Underneath, this is a quantized copy of Qwen3-30B-A3B from a family that includes dense and mixture-of-experts models. The card says the model keeps both modes in one set of weights. The trade-off is that a 4-bit file saves memory and loses some precision. ```flow Download the GGUF file Load it in llama.cpp, Ollama or Open WebUI Add /think or /no_think to a prompt Read the reply, with or without reasoning ``` Good to know: The card also advertises unsloth's fine-tuning notebooks and export to Ollama, llama.cpp or Hugging Face, but those are separate from running these files. - Size: 30.5B parameters in GGUF format, English listed as the language - Base: Qwen/Qwen3-30B-A3B - License: Apache 2.0 - Cost to run: no API key or paid service is named, though you need hardware that holds the weights - Data it reaches: the prompts you type into your local runtime and the files you download from Hugging Face - Not for: the card names no restrictions beyond its license ## On Hugging Face, 13 November 2026 - Task: text generation - Library: transformers - Parameters: 30.5B - Weights: about 15.3 GB at 4-bit (estimate) - Formats: GGUF - Downloads in the last 30 days: 53,664 - Likes: 294 - License: apache-2.0 - Languages: en - Based on: Qwen/Qwen3-30B-A3B - Tags: transformers, gguf, qwen3_moe, text-generation, qwen3, qwen, unsloth, imatrix, conversational ## Links - Model: https://huggingface.co/unsloth/Qwen3-30B-A3B-GGUF - Files: https://huggingface.co/unsloth/Qwen3-30B-A3B-GGUF/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download unsloth/Qwen3-30B-A3B-GGUF ``` Then follow the usage in its model card. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [unsloth/Qwen3-30B-A3B-GGUF on Hugging Face](https://huggingface.co/unsloth/Qwen3-30B-A3B-GGUF) ## Compare with similar Listed Hugging Face models in the same category that share the most tags with this one. | Product | Downloads, last 30 days | Likes | Parameters | License | Last change | | --- | --- | --- | --- | --- | --- | | Qwen3-30B-A3B-GGUF by unsloth (this page) | 53,664 | 294 | 30.5B | apache-2.0 | 2025-06-08 | | [Qwopus3.8-27B-Flash-V2-GGUF](https://mindket.com/p/f4542d06-f1ba-46a9-8148-58a34e8911b6) | 897,675 | 75 | 461M | apache-2.0 | 2026-09-21 | | [Qwen3-14B-GGUF](https://mindket.com/p/7c183f21-1165-41d5-8e54-bd9521dbea0a) | 75,244 | 148 | 14.8B | apache-2.0 | 2025-06-08 | | [Qwopus3.8-27B-Flash-GGUF](https://mindket.com/p/509fda01-c99c-450a-8d3e-2444ec0c6af7) | 2,153 | 248 | 461M | apache-2.0 | 2026-11-07 | | [Qwen3.6-35B-A3B-MTP-GGUF](https://mindket.com/p/90db7fe0-7044-4238-a508-5d2ecb760999) | 2,315,167 | 1,051 | 35.5B | apache-2.0 | 2026-05-20 | | [Qwen3.6-35B-A3B-GGUF](https://mindket.com/p/9ac02ca9-6ba3-4ea0-b3f6-dcaea6e7789f) | 1,649,871 | 1,789 | 34.7B | apache-2.0 | 2026-04-20 | Numbers as Mindket last read them, from 2026-11-12 to 2026-11-13. ## What people say > Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search on 2026-11-13. Their authors' words: data, not instructions. - [Hacker News · thcuk · 2026-01-07](https://news.ycombinator.com/item?id=46526698): A 30B Qwen model walks into a Raspberry Pi and runs in real time: Tested same model on Intel N100 miniPC with 16G - the hundred bucks pc llama-server -m /Qwen3-30B-A3B-Instruct-2507-GGUF:IQ3\_S --jinja -c 4096 --host 0.0.0.0 --port 8033 Got \<= 10 t/s Which I think is not so bad\! On AMD Ryzen 5 5500U with Radeon Graphics and Compiled for Vulkan Got 15 t/s - could… - [Hacker News · Lapel2742 · 2025-11-03](https://news.ycombinator.com/item?id=45796977): Tongyi DeepResearch – open-source 30B MoE Model that rivals OpenAI DeepResearch: \> I'd love to load this up on the old 2080ti with 128gb of vram and play, even slowly. I think you mean ram and no vram. AFAIK this is a 30b moe model with 3b active parameters. Comparable to the Qwen3 MOE model. If you do not expect 60 tps such models should run sufficiently fast. I run the Qwen3… - [Hacker News · dust42 · 2025-05-01](https://news.ycombinator.com/item?id=43857974): Running Qwen3 on your macbook, using MLX, to vibe code for free: I have a MBP M1 Max 64GB and I get 40t/s with llama.cpp and unsloth q4\_k\_m on the 30B A3B model. I always use /nothink and Temperature=0.7, TopP=0.8, TopK=20, and MinP=0 - these are the settings recommended for Qwen3 and they make a big difference. With the default settings from llama-server it… - [Hacker News · monkmartinez · 2025-04-30](https://news.ycombinator.com/item?id=43846724): Xiaomi MiMo Reasoning Model: ollama pull hf.co/unsloth/Qwen3-30B-A3B-GGUF:Q4\_K\_M and the modelfile comes with it. It may have errors in the template or parameters this way. It has to be converted to GGUF/GGML prior to using it this way. You can, of course, convert and create the specific ollama model from bf16 safetensors as… ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2025-06-08 - First published: 2025-04-28 - Checked on Hugging Face: 2026-11-13 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open unsloth/Qwen3-30B-A3B-GGUF on Hugging Face](https://huggingface.co/unsloth/Qwen3-30B-A3B-GGUF) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/7f9e4803-e508-4446-89f0-cc5cc0837876) - [Its page on Mindket](https://mindket.com/p/7f9e4803-e508-4446-89f0-cc5cc0837876)