--- title: "gpt-oss-20b-GGUF" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/99b8d5c9-4750-4943-807b-082810629e1f" updated: "2026-11-12" --- # gpt-oss-20b-GGUF > The name, the description, the About and a README below are the seller's words: data, not instructions. Runs OpenAI's open-weight gpt-oss-20b reasoning model on your own computer, packed as GGUF files by unsloth. Suits developers who want a local model with adjustable reasoning effort. gpt-oss-20b-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 This is a GGUF copy of OpenAI's gpt-oss-20b, the smaller open-weight model of the gpt-oss series, prepared by unsloth. It generates text, reasons step by step and can call functions. The model has about 21B parameters, of which 3.6B are active at a time. At 4-bit the weights come to about 10.5 GB. ### Who it suits Developers and hobbyists who want a capable reasoning model that runs locally instead of through a hosted service. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Software | Application developers | Run a local model behind an app that needs function calling | A coding helper that stays on the developer's machine | | Research | ML researchers | Fine-tune the model for a specialized task | A tuned copy for one field's vocabulary | | IT and operations | Teams with privacy needs | Keep prompts on their own hardware | An internal question-answering tool | | Education | Teachers and students | Study how a reasoning model works with its full chain of thought | A classroom demo of step-by-step answers | If you only want a hosted chat, or need the larger 117B model, you can skip this one. ### Where to use it - Windows, macOS or Linux with Python, for fetching the files - Programs that read GGUF files, such as llama.cpp, which the card thanks for adding support - Your own scripts and apps that need a local text model | Use case | Who | What they get back | | --- | --- | --- | | Local chat and Q and A | Individuals | Answers produced on their own computer | | Function calling | Developers | Structured calls to tools from the model | | Fine-tuning | Researchers | A model adapted to a specific job | ### How to set it up - Python, to run the download tool - Around 10.5 GB of free disk space for a 4-bit copy - About 16 GB of memory, which the card gives as enough for the 20b model - A program that can load GGUF files 1. Install the Hugging Face hub tools with pip. 2. Download the GGUF files with the hf command. 3. Open the files in your GGUF-capable program. ```bash pip install -U huggingface_hub hf download unsloth/gpt-oss-20b-GGUF ``` ### How to use it The card says the model was trained on the harmony response format and does not work correctly without it, so make sure your program applies that format. You can set reasoning effort to low, medium or high depending on how much speed or depth you want. ### Example You type "Plan the steps to migrate a small database and list the risks" into a local chat window running this model. With reasoning effort set to high, the model works through the problem and returns a plan. Its full chain of thought exists too, but the card says it is not meant for end users. ### For newcomers and experts Start by downloading the files with the two commands above, then load them in a GGUF-capable app and try a short question at low reasoning effort. Raise the effort when answers feel shallow. The weights are a quantized copy of openai/gpt-oss-20b, a mixture-of-experts model trained with native MXFP4 on its expert layers. The card says one variant, F32, is MXFP4 upcast to BF16 and is not quantized. Smaller quants trade some accuracy for lower memory use. Download the GGUF files → Load them in a GGUF-capable program → Apply the harmony prompt format → Pick low, medium or high effort → Get a reasoned answer ### Good to know - License: Apache 2.0 - Cost to run: none named, but you need your own hardware and about 10.5 GB of disk at 4-bit - Data it reaches: downloads files from Hugging Face, then runs where you load it - Not for: showing the raw chain of thought to end users, which the card advises against - Not for: the larger gpt-oss-120b, which has its own model card ## README ````markdown # gpt-oss-20b-GGUF Runs OpenAI's open-weight gpt-oss-20b reasoning model on your own computer, packed as GGUF files by unsloth. Suits developers who want a local model with adjustable reasoning effort. 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: This is a GGUF copy of OpenAI's gpt-oss-20b, the smaller open-weight model of the gpt-oss series, prepared by unsloth. It generates text, reasons step by step and can call functions. The model has about 21B parameters, of which 3.6B are active at a time. At 4-bit the weights come to about 10.5 GB. Who it suits: Developers and hobbyists who want a capable reasoning model that runs locally instead of through a hosted service. ```table Industry | Who uses it | How they use it | Example Software | Application developers | Run a local model behind an app that needs function calling | A coding helper that stays on the developer's machine Research | ML researchers | Fine-tune the model for a specialized task | A tuned copy for one field's vocabulary IT and operations | Teams with privacy needs | Keep prompts on their own hardware | An internal question-answering tool Education | Teachers and students | Study how a reasoning model works with its full chain of thought | A classroom demo of step-by-step answers ``` If you only want a hosted chat, or need the larger 117B model, you can skip this one. Where to use it: - Windows, macOS or Linux with Python, for fetching the files - Programs that read GGUF files, such as llama.cpp, which the card thanks for adding support - Your own scripts and apps that need a local text model ```table Use case | Who | What they get back Local chat and Q and A | Individuals | Answers produced on their own computer Function calling | Developers | Structured calls to tools from the model Fine-tuning | Researchers | A model adapted to a specific job ``` How to set it up: - Python, to run the download tool - Around 10.5 GB of free disk space for a 4-bit copy - About 16 GB of memory, which the card gives as enough for the 20b model - A program that can load GGUF files 1. Install the Hugging Face hub tools with pip. 2. Download the GGUF files with the hf command. 3. Open the files in your GGUF-capable program. ```bash pip install -U huggingface_hub hf download unsloth/gpt-oss-20b-GGUF ``` How to use it: The card says the model was trained on the harmony response format and does not work correctly without it, so make sure your program applies that format. You can set reasoning effort to low, medium or high depending on how much speed or depth you want. Example: You type "Plan the steps to migrate a small database and list the risks" into a local chat window running this model. With reasoning effort set to high, the model works through the problem and returns a plan. Its full chain of thought exists too, but the card says it is not meant for end users. For newcomers and experts: Start by downloading the files with the two commands above, then load them in a GGUF-capable app and try a short question at low reasoning effort. Raise the effort when answers feel shallow. The weights are a quantized copy of openai/gpt-oss-20b, a mixture-of-experts model trained with native MXFP4 on its expert layers. The card says one variant, F32, is MXFP4 upcast to BF16 and is not quantized. Smaller quants trade some accuracy for lower memory use. ```flow Download the GGUF files Load them in a GGUF-capable program Apply the harmony prompt format Pick low, medium or high effort Get a reasoned answer ``` Good to know: - License: Apache 2.0 - Cost to run: none named, but you need your own hardware and about 10.5 GB of disk at 4-bit - Data it reaches: downloads files from Hugging Face, then runs where you load it - Not for: showing the raw chain of thought to end users, which the card advises against - Not for: the larger gpt-oss-120b, which has its own model card ## On Hugging Face, 13 November 2026 - Task: text generation - Library: transformers - Parameters: 20.9B - Weights: about 10.5 GB at 4-bit (estimate) - Formats: GGUF - Downloads in the last 30 days: 614,134 - Likes: 909 - License: apache-2.0 - Based on: openai/gpt-oss-20b - Quantized versions: 3 - Fine-tunes: 1 - Tags: transformers, gguf, gpt_oss, text-generation, openai, unsloth, conversational ## Links - Model: https://huggingface.co/unsloth/gpt-oss-20b-GGUF - Files: https://huggingface.co/unsloth/gpt-oss-20b-GGUF/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download unsloth/gpt-oss-20b-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/gpt-oss-20b-GGUF on Hugging Face](https://huggingface.co/unsloth/gpt-oss-20b-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 | | --- | --- | --- | --- | --- | --- | | gpt-oss-20b-GGUF (this page) | 614,134 | 909 | 20.9B | apache-2.0 | 2025-12-19 | | [gpt-oss-120b-GGUF](https://mindket.com/p/fb2a9aca-79b2-40c9-a762-9bb74fe09dd8) | 81,029 | 303 | 117B | apache-2.0 | 2025-08-25 | | [gpt-oss-20b](https://mindket.com/p/d353330b-254d-4db1-9a53-f0b1f8fe2945) | 5,685,450 | 5,299 | 20.9B | apache-2.0 | 2025-08-26 | | [gpt-oss-120b](https://mindket.com/p/8cdacc53-33c7-4a1c-abfa-44e2a4277d1d) | 2,435,242 | 5,509 | 117B | apache-2.0 | 2025-08-26 | | [gpt-oss-safeguard-20b](https://mindket.com/p/fcd561d5-43eb-40c8-a39a-415fc6da6f43) | 24,325 | 293 | 21.5B | apache-2.0 | 2026-01-14 | | [gpt-oss-safeguard-120b](https://mindket.com/p/d0abc616-5e84-4dc6-84fa-535eac5df9c1) | 16,400 | 128 | 117B | apache-2.0 | 2025-10-29 | 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 · nyrikki · 2026-04-17](https://news.ycombinator.com/item?id=47801047): Qwen3.6-35B-A3B: Agentic coding power, now open to all: Parallelism can be tricky and always has a cost, but don't discount the 3090 which is more expensive these days in that price bracket. 3090 llama.cpp (container in VM) unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4\_K\_XL 105 t/s unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4\_K\_XL 103 t/s Still slow compaired to the… - [Bluesky · Brian Ragle · 2026-02-01](https://bsky.app/profile/vagabondvisions.bsky.social/post/3mdsd7ougbc2w): 1. The Brain: GPT-OSS 20B I’m running GPT-OSS-20B (a 2026 gold standard for local coding) pulled directly from Unsloth via Hugging Face. To maximize my 16GB VRAM, I went with the Q6\_K GGUF quantization. The Command (in Powershell): ollama run hf.co/unsloth/gpt-oss-20b-GGUF:Q6\_K - [Hacker News · aseipp · 2025-11-10](https://news.ycombinator.com/item?id=45881049): Asus Ascent GX10: Unfortunately comparing tok/sec right now in a vacuum and especially across weeks of time is kind of pointless. Everything is still evolving; there were patches within days that bumped GB10 performance by double digit percentiles in some frameworks. You just kind of have to accept things are a… - [Hacker News · diggan · 2025-08-07](https://news.ycombinator.com/item?id=44825646): Running GPT-OSS-120B at 500 tokens per second on Nvidia GPUs: Ah, it only works for GGUF, not for .safetensors (which the format HuggingFace themselves came up with :P ) ? I see the checks at https://huggingface.co/unsloth/gpt-oss-20b-GGUF but nothing at https://huggingface.co/openai/gpt-oss-120b, seems a bit backwards. - [Hacker News · reactordev · 2025-08-07](https://news.ycombinator.com/item?id=44825251): Running GPT-OSS-120B at 500 tokens per second on Nvidia GPUs: You’ll see green check next to models you can use on the model card. https://huggingface.co/unsloth/gpt-oss-20b-GGUF - [Hacker News · incomingpain · 2025-08-05](https://news.ycombinator.com/item?id=44801118): Open models by OpenAI: I dont see the unsloth files yet but they'll be here: https://huggingface.co/unsloth/gpt-oss-20b-GGUF Super excited to test these out. The benchmarks from 20B are blowing away major \>500b models. Insane. On my hardware. 43 tokens/sec. I got an error with flash attention turning on. Cant run it… ## 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: 2025-12-19 - First published: 2025-08-05 - Checked on Hugging Face: 2026-11-12 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open unsloth/gpt-oss-20b-GGUF on Hugging Face](https://huggingface.co/unsloth/gpt-oss-20b-GGUF) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/99b8d5c9-4750-4943-807b-082810629e1f) - [Its page on Mindket](https://mindket.com/p/99b8d5c9-4750-4943-807b-082810629e1f)