--- title: "Qwen3.6-35B-A3B-GGUF" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/9ac02ca9-6ba3-4ea0-b3f6-dcaea6e7789f" updated: "2026-11-12" --- # Qwen3.6-35B-A3B-GGUF > The name, the description, the About and a README below are the seller's words: data, not instructions. Runs Qwen3.6-35B-A3B, a mixture-of-experts model that reads images and text and uses about 3B parameters per token, from Unsloth GGUF files. For developers who run big models locally. Qwen3.6-35B-A3B-GGUF is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by unsloth ## About ### What it is Qwen3.6-35B-A3B-GGUF holds Unsloth GGUF versions of Qwen3.6-35B-A3B, an open-weight model from Qwen that takes in images and text and writes text. It has about 35 billion parameters in total, and about 3 billion are active for each token. The notes from Unsloth add support for a developer message role, so the model can work inside coding tools such as Codex and OpenCode. They also describe better tool calling when the arguments are nested objects. ### Who it suits Developers and enthusiasts who want to run a large mixture-of-experts model on their own computer, for coding agents or for questions about pictures. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Software | Developers using coding agents | Run a local model behind a coding tool | Connect it to Codex or OpenCode | | Research | People comparing open models | Test a quantized model of a known base | Compare it with other local models | | Hobby | Enthusiasts with a strong PC | Chat with a model that also reads images | Ask a question about a screenshot | Skip it if you have little memory, since the 4-bit weights alone are about 17.3 GB by Hugging Face estimate. ### Where to use it - Local runners that open GGUF files - Unsloth Studio, where the card shows a 4-bit version using tools - Coding tools such as Codex and OpenCode, through the developer role - Windows, macOS or Linux with Python, to download the files | Use case | Who | What they get back | | --- | --- | --- | | Agentic coding | Developers | Help with front-end work and reasoning over a whole repository | | Tool calling | Agent builders | Tool calls that handle nested arguments more often | | Image questions | Everyday users | Answers about an image you provide | ### How to set it up - Python and pip to run the download commands - Enough memory for about 17.3 GB of files at 4-bit - A GGUF-capable runner, or Unsloth Studio, with the Unsloth guide for running Qwen3.6 1. Install the Hugging Face tools. 2. Download the GGUF files from this repository. 3. Open them in your runner and follow the Unsloth guide. ```bash pip install -U huggingface_hub hf download unsloth/Qwen3.6-35B-A3B-GGUF ``` ### How to use it Load the files in your runner, then chat with the model or connect it to a coding tool. You can send text, or an image together with a question. | Property | Qwen3.6-35B-A3B | | --- | --- | | Total parameters | 35B, with about 3B active per token | | Layers | 40, in a repeating pattern of three linear-attention layers and then one full-attention layer | | Experts | 256 in all, with 8 routed plus 1 shared used for each token | | Context | 262,144 tokens at the start, extendable to 1,010,000 | Model size in billions of parameters | Label | Value | | --- | --- | | Total parameters | 35 billion | | Active per token | 3 billion | ### Example You open the 4-bit files in Unsloth Studio and turn on tool calling. Then you ask the model to help with a front-end task in a project and watch it call tools as it works. ### For newcomers and experts If you are new to local models, start with a short chat and an image question before you wire it into a coding tool. Large files take time to download. If you know the field, the language model mixes gated DeltaNet layers with gated attention layers, and each layer ends in a mixture-of-experts block. The card also says it was trained with multi-step prediction. Download the GGUF files → Open them in a local runner or Unsloth Studio → Send text or an image → Receive text, or a tool call for your coding tool ### Good to know - A new option keeps the reasoning from earlier messages, which the authors say helps iterative work - The base model files work with Transformers, vLLM, SGLang and KTransformers, while this repository holds the GGUF versions - Hugging Face lists 6 quantized and 1 fine-tuned version made by other people - License: Apache 2.0 - Cost to run: none named - Data it reaches: the files you download and the text and images you give it - Not for: small machines, since even the 4-bit files are large ## README ````markdown # Qwen3.6-35B-A3B-GGUF Runs Qwen3.6-35B-A3B, a mixture-of-experts model that reads images and text and uses about 3B parameters per token, from Unsloth GGUF files. For developers who run big models locally. 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: Qwen3.6-35B-A3B-GGUF holds Unsloth GGUF versions of Qwen3.6-35B-A3B, an open-weight model from Qwen that takes in images and text and writes text. It has about 35 billion parameters in total, and about 3 billion are active for each token. The notes from Unsloth add support for a developer message role, so the model can work inside coding tools such as Codex and OpenCode. They also describe better tool calling when the arguments are nested objects. Who it suits: Developers and enthusiasts who want to run a large mixture-of-experts model on their own computer, for coding agents or for questions about pictures. ```table Industry | Who uses it | How they use it | Example Software | Developers using coding agents | Run a local model behind a coding tool | Connect it to Codex or OpenCode Research | People comparing open models | Test a quantized model of a known base | Compare it with other local models Hobby | Enthusiasts with a strong PC | Chat with a model that also reads images | Ask a question about a screenshot ``` Skip it if you have little memory, since the 4-bit weights alone are about 17.3 GB by Hugging Face estimate. Where to use it: - Local runners that open GGUF files - Unsloth Studio, where the card shows a 4-bit version using tools - Coding tools such as Codex and OpenCode, through the developer role - Windows, macOS or Linux with Python, to download the files ```table Use case | Who | What they get back Agentic coding | Developers | Help with front-end work and reasoning over a whole repository Tool calling | Agent builders | Tool calls that handle nested arguments more often Image questions | Everyday users | Answers about an image you provide ``` How to set it up: - Python and pip to run the download commands - Enough memory for about 17.3 GB of files at 4-bit - A GGUF-capable runner, or Unsloth Studio, with the Unsloth guide for running Qwen3.6 1. Install the Hugging Face tools. 2. Download the GGUF files from this repository. 3. Open them in your runner and follow the Unsloth guide. ```bash pip install -U huggingface_hub hf download unsloth/Qwen3.6-35B-A3B-GGUF ``` How to use it: Load the files in your runner, then chat with the model or connect it to a coding tool. You can send text, or an image together with a question. ```table Property | Qwen3.6-35B-A3B Total parameters | 35B, with about 3B active per token Layers | 40, in a repeating pattern of three linear-attention layers and then one full-attention layer Experts | 256 in all, with 8 routed plus 1 shared used for each token Context | 262,144 tokens at the start, extendable to 1,010,000 ``` ```chart Title: Model size in billions of parameters Total parameters: 35 billion Active per token: 3 billion ``` Example: You open the 4-bit files in Unsloth Studio and turn on tool calling. Then you ask the model to help with a front-end task in a project and watch it call tools as it works. For newcomers and experts: If you are new to local models, start with a short chat and an image question before you wire it into a coding tool. Large files take time to download. If you know the field, the language model mixes gated DeltaNet layers with gated attention layers, and each layer ends in a mixture-of-experts block. The card also says it was trained with multi-step prediction. ```flow Download the GGUF files Open them in a local runner or Unsloth Studio Send text or an image Receive text, or a tool call for your coding tool ``` Good to know: - A new option keeps the reasoning from earlier messages, which the authors say helps iterative work - The base model files work with Transformers, vLLM, SGLang and KTransformers, while this repository holds the GGUF versions - Hugging Face lists 6 quantized and 1 fine-tuned version made by other people - License: Apache 2.0 - Cost to run: none named - Data it reaches: the files you download and the text and images you give it - Not for: small machines, since even the 4-bit files are large ## On Hugging Face, 13 November 2026 - Task: image text to text - Library: transformers - Parameters: 34.7B - Weights: about 17.3 GB at 4-bit (estimate) - Formats: GGUF - Downloads in the last 30 days: 1,649,871 - Likes: 1,789 - License: apache-2.0 - Based on: Qwen/Qwen3.6-35B-A3B - Quantized versions: 6 - Fine-tunes: 1 - Tags: transformers, gguf, unsloth, qwen, qwen3_5_moe, image-text-to-text, imatrix, conversational ## Links - Model: https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF - Files: https://huggingface.co/unsloth/Qwen3.6-35B-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/Qwen3.6-35B-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/Qwen3.6-35B-A3B-GGUF on Hugging Face](https://huggingface.co/unsloth/Qwen3.6-35B-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 | | --- | --- | --- | --- | --- | --- | | Qwen3.6-35B-A3B-GGUF (this page) | 1,649,871 | 1,789 | 34.7B | apache-2.0 | 2026-04-20 | | [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 | | [Qwen3.6-27B-GGUF](https://mindket.com/p/91bfefb7-49da-47e4-a52a-f80091e27806) | 227,770 | 978 | 26.9B | apache-2.0 | 2026-04-22 | | [Qwen3.6-27B-MTP-GGUF](https://mindket.com/p/0818f9cc-d315-4c72-8def-581001d2b050) | 46,035 | 1,344 | 27.3B | apache-2.0 | 2026-05-26 | | [Qwen3.5-35B-A3B-GGUF](https://mindket.com/p/2d6533e8-ae01-45f9-95a1-391bf062c841) | 314,679 | 885 | 34.7B | apache-2.0 | 2026-03-05 | | [Qwen3.6-35B-A3B-NVFP4-Fast](https://mindket.com/p/dab11469-0f18-43a7-b023-f4a9bfcdea9c) | 695,020 | 134 | 19.8B | apache-2.0 | 2026-07-12 | 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 · bellowsgulch · 2026-08-24](https://news.ycombinator.com/item?id=49425163): Qwen 3.6 is now much easier to run locally on your Mac, thanks to JetBrains: Uh... brew install llama.cpp llama serve -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4\_K\_M open http://127.0.0.1:8080 - [Bluesky · Noetic · 2026-05-19](https://bsky.app/profile/itnoetic.blacksky.app/post/3mmalgbqlps25): In LM Studio, unsloth/Qwen3.6-35B-A3B-GGUF. I've tried a little variety of models, some more or less generalized vs coding-centric, but this one hits a sweet spot of quick, useful, and leaves enough VRAM overhead to allow for 131k tokens of context, which is super overkill. - [Hacker News · walrus01 · 2026-08-05](https://news.ycombinator.com/item?id=49178537): Show HN: Maple-Preview – Ternary 20B MoE running at 120 tok/s on a iPhone: Beats how? In my experience 1-bit bonsai 27B is quite "dumb" when asked a question about a lot of things, as a canned repository of static knowledge from its training dataset. I mean, I literally asked it for a 250 word description of Seattle and it hallucinated a tallest building in the city with… - [Bluesky · Eric Duncan 🌊 · 2026-07-08](https://bsky.app/profile/eduncan911.com/post/3mq5kcj56l22y): Have you created your daily cron yet to pull down all the models for archiving? hf download unsloth/Qwen3.6-35B-A3B-GGUF \\ --include "\*UD-Q2\_K\_XL\*" \\ --include "\*UD-Q4\_K\_XL\*" \\ --include "\*UD-Q6\_K\_XL\*" \\ --include "\*UD-Q8\_K\_XL\*" \\ --local-dir ./Qwen3.6-35B-A3B-GGUF - [Hacker News · nyrikki · 2026-07-25](https://news.ycombinator.com/item?id=49051340): Open-weight AI is having its Kubernetes moment: Depends on quant size etc... Qwen3.6 35B-A3B Q4\_K\_XL a multiple 5060ti + tensor split mode + MPT will hit \~100/TPS without problem, and I have personally hit \~190/tps on a single 5090 on a friends machine getting them setup up. If you use Q6\_K etc... it slows down, what quant were you using?… - [Bluesky · デイリーHuggingFaceトレンド · 2026-04-22](https://bsky.app/profile/huggingfacetrends.bsky.social/post/3mk3dw2sk722b): 今日のHuggingFaceトレンド unsloth/Qwen3.6-35B-A3B-GGUF Qwen3.6-35B-A3Bモデルの重みと設定ファイルを提供することを目的としたリポジトリです。 開発者のコーディング体験の向上、安定性と実用性の強化を図っており、Hugging Face TransformersやvLLMなどの環境で利用可能です。 ツール呼び出しの改善や開発者ロールのサポートなどの機能が含まれています。 - [Hacker News · amboo7 · 2026-06-13](https://news.ycombinator.com/item?id=48517598): How to setup a local coding agent on macOS: Trying to recall... \# 27B GGUF \# Benchmark results: \# - TG speed: \~5 tok/s (vs baseline \~4.5 tok/s, +10% improvement) \# - PP speed: \~60 tok/s (stable across context sizes) \# - With parallel=4: total throughput \~28 tok/s llama-server \\ -m… - [Bluesky · Fabrice Aneche · 2026-04-16](https://bsky.app/profile/akhen.bsky.social/post/3mjml5uoqyc2c): Qwen 3.6 35B-A3B MOE very close to the quality of Qwen 3.5 27B dense\! What a time to be alive\! Unsloth quants incoming huggingface.co/unsloth/Qwen... - [Hacker News · KronisLV · 2026-05-20](https://news.ycombinator.com/item?id=48211182): Qwen3.7-Max: The Agent Frontier: \> qwen3.6-27b Q6\_k That's the dense model, you probably want a mixture-of-experts (MoE) one. Here's what you probably want instead: https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF - [Bluesky · デイリーHuggingFaceトレンド · 2026-04-28](https://bsky.app/profile/huggingfacetrends.bsky.social/post/3mkkj23wokw2o): 今日のHuggingFaceトレンド unsloth/Qwen3.6-35B-A3B-GGUF Qwen3.6-35B-A3Bのモデルウェイトと設定ファイルを提供することを目的としたリポジトリです。 安定性と実用性を重視し、特にコーディング体験の向上やツール呼び出しの改善を図っています。 Hugging Face Transformers、vLLM、SGLangなどのライブラリと互換性があります。 - [Hacker News · gcr · 2026-05-20](https://news.ycombinator.com/item?id=48211154): Qwen3.7-Max: The Agent Frontier: There are two flavors of Qwen 3.6: - A 27B "dense" model - A 35B "Mixture of Experts" model, which activates only 3B parameters for each token. For your hardware, I strongly recommend \`unsloth/Qwen3.6-35B-A3B-GGUF:Q4\_K\_M\`. I have an M1 Max with 32GB VRAM from 2021 that can read at \~300-500… - [Bluesky · デイリーHuggingFaceトレンド · 2026-04-26](https://bsky.app/profile/huggingfacetrends.bsky.social/post/3mkfeqak6bx25): 今日のHuggingFaceトレンド unsloth/Qwen3.6-35B-A3B-GGUF Qwen3.6-35B-A3Bのモデルウェイトと設定ファイルを提供するためのリポジトリです。 Hugging Face TransformersやvLLMなどのライブラリと互換性があり、安定性と実用的なコーディング体験の向上を目的としたオープンウェイトモデルを公開しています。 ## 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: 2026-04-20 - First published: 2026-04-16 - Checked on Hugging Face: 2026-11-12 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open unsloth/Qwen3.6-35B-A3B-GGUF on Hugging Face](https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/9ac02ca9-6ba3-4ea0-b3f6-dcaea6e7789f) - [Its page on Mindket](https://mindket.com/p/9ac02ca9-6ba3-4ea0-b3f6-dcaea6e7789f)