Hugging Face's logo

bartowski
/
openai_gpt-oss-20b-GGUF

Text Generation
GGUF
conversational
Model card Files
xet
Community
1
openai_gpt-oss-20b-GGUF / README.md
bartowski's picture
bartowski
Update README.md
e39ba3a verified
preview code
|
raw
history blame contribute delete
14.4 kB
 ---
quantized_by: bartowski
pipeline_tag: text-generation
base_model_relation: quantized
base_model: openai/gpt-oss-20b
---
## Llamacpp imatrix Quantizations of gpt-oss-20b by openai
Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/b6096">b6096</a> for quantization.
Original model: https://huggingface.co/openai/gpt-oss-20b
All quants made using imatrix option with combined_all_medium dataset from [Ed Addario here](https://huggingface.co/datasets/eaddario/imatrix-calibration/blob/main/combined_all_medium.parquet)
Run them in [LM Studio](https://lmstudio.ai/)
Run them directly with [llama.cpp](https://github.com/ggerganov/llama.cpp), or any other llama.cpp based project
All quants keep the feed forward networks at mxfp4 for optimal performance, which does mean the size differences are negligible unfortunately, but being provided just because.
## Prompt format
No chat template specified so default is used. This may be incorrect, check original model card for details.
```
<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-08-05
Reasoning: medium
# Valid channels: analysis, commentary, final. Channel must be included for every message.<|end|><|start|>developer<|message|># Instructions
{system_prompt}<|end|><|start|>user<|message|>{prompt}<|end|><|start|>assistant
```
## Download a file (not the whole branch) from below:
Use this one:
| Filename | Quant type | File Size | Split | Description |
| -------- | ---------- | --------- | ----- | ----------- |
| [gpt-oss-20b-MXFP4.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-MXFP4.gguf) | MXFP4 | 12.1GB | false | Full MXFP4 weights, *recommended* for this model. |
The reason is, the FFN (feed forward networks) of gpt-oss do not behave nicely when quantized to anything other than MXFP4, so they are kept at that level for everything.
The rest of these are provided for your own interest in case you feel like experimenting, but the size savings is basically non-existent so I would not recommend running them, they are provided simply for show:
| Filename | Quant type | File Size | Split | Description |
| -------- | ---------- | --------- | ----- | ----------- |
| [gpt-oss-20b-Q6_K_L.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q6_K_L.gguf) | Q6_K_L | 12.04GB | false | Uses Q8_0 for embed and output weights. Q6_K with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q6_K.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q6_K.gguf) | Q6_K | 12.04GB | false | Q6_K with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q5_K_L.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q5_K_L.gguf) | Q5_K_L | 11.91GB | false | Uses Q8_0 for embed and output weights. Q5_K with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q4_K_L.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q4_K_L.gguf) | Q4_K_L | 11.89GB | false | Uses Q8_0 for embed and output weights. Q4_K with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q2_K_L.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q2_K_L.gguf) | Q2_K_L | 11.85GB | false | Uses Q8_0 for embed and output weights. Q2_K with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q3_K_XL.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q3_K_XL.gguf) | Q3_K_XL | 11.78GB | false | Uses Q8_0 for embed and output weights. Q3_K_L with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q5_K_M.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q5_K_M.gguf) | Q5_K_M | 11.73GB | false | Q5_K_M with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q5_K_S.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q5_K_S.gguf) | Q5_K_S | 11.72GB | false | Q5_K_S with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q4_K_M.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q4_K_M.gguf) | Q4_K_M | 11.67GB | false | Q4_K_M with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q4_K_S.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q4_K_S.gguf) | Q4_K_S | 11.67GB | false | Q4_K_S with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q4_1.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q4_1.gguf) | Q4_1 | 11.59GB | false | Q4_1 with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ4_NL.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ4_NL.gguf) | IQ4_NL | 11.56GB | false | IQ4_NL with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ4_XS.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ4_XS.gguf) | IQ4_XS | 11.56GB | false | IQ4_XS with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q3_K_M.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q3_K_M.gguf) | Q3_K_M | 11.56GB | false | Q3_K_M with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ3_M.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ3_M.gguf) | IQ3_M | 11.56GB | false | IQ3_M with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ3_XS.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ3_XS.gguf) | IQ3_XS | 11.56GB | false | IQ3_XS with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ3_XXS.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ3_XXS.gguf) | IQ3_XXS | 11.56GB | false | IQ3_XXS with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q2_K.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q2_K.gguf) | Q2_K | 11.56GB | false | Q2_K with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q3_K_S.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q3_K_S.gguf) | Q3_K_S | 11.55GB | false | Q3_K_S with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ2_M.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ2_M.gguf) | IQ2_M | 11.55GB | false | IQ2_M with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ2_S.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ2_S.gguf) | IQ2_S | 11.55GB | false | IQ2_S with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q4_0.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q4_0.gguf) | Q4_0 | 11.52GB | false | Q4_0 with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ2_XS.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ2_XS.gguf) | IQ2_XS | 11.51GB | false | IQ2_XS with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-IQ2_XXS.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-IQ2_XXS.gguf) | IQ2_XXS | 11.51GB | false | IQ2_XXS with all FFN kept at MXFP4_MOE. |
| [gpt-oss-20b-Q3_K_L.gguf](https://huggingface.co/bartowski/openai_gpt-oss-20b-GGUF/blob/main/openai_gpt-oss-20b-Q3_K_L.gguf) | Q3_K_L | 11.49GB | false | Q3_K_L with all FFN kept at MXFP4_MOE. |
## Embed/output weights
Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.
## Downloading using huggingface-cli
<details>
<summary>Click to view download instructions</summary>
First, make sure you have hugginface-cli installed:
```
pip install -U "huggingface_hub[cli]"
```
Then, you can target the specific file you want:
```
huggingface-cli download bartowski/openai_gpt-oss-20b-GGUF --include "openai_gpt-oss-20b-Q4_K_M.gguf" --local-dir ./
```
If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:
```
huggingface-cli download bartowski/openai_gpt-oss-20b-GGUF --include "openai_gpt-oss-20b-Q8_0/*" --local-dir ./
```
You can either specify a new local-dir (openai_gpt-oss-20b-Q8_0) or download them all in place (./)
</details>
## ARM/AVX information
Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass.
Now, however, there is something called "online repacking" for weights. details in [this PR](https://github.com/ggerganov/llama.cpp/pull/9921). If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.
As of llama.cpp build [b4282](https://github.com/ggerganov/llama.cpp/releases/tag/b4282) you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.
Additionally, if you want to get slightly better quality for , you can use IQ4_NL thanks to [this PR](https://github.com/ggerganov/llama.cpp/pull/10541) which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase.
<details>
<summary>Click to view Q4_0_X_X information (deprecated</summary>
I'm keeping this section to show the potential theoretical uplift in performance from using the Q4_0 with online repacking.
<details>
<summary>Click to view benchmarks on an AVX2 system (EPYC7702)</summary>
| model | size | params | backend | threads | test | t/s | % (vs Q4_0) |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | ------------: | -------------------: |-------------: |
| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp512 | 204.03 ± 1.03 | 100% |
| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp1024 | 282.92 ± 0.19 | 100% |
| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp2048 | 259.49 ± 0.44 | 100% |
| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg128 | 39.12 ± 0.27 | 100% |
| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg256 | 39.31 ± 0.69 | 100% |
| qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg512 | 40.52 ± 0.03 | 100% |
| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp512 | 301.02 ± 1.74 | 147% |
| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp1024 | 287.23 ± 0.20 | 101% |
| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp2048 | 262.77 ± 1.81 | 101% |
| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg128 | 18.80 ± 0.99 | 48% |
| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg256 | 24.46 ± 3.04 | 83% |
| qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg512 | 36.32 ± 3.59 | 90% |
| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | pp512 | 271.71 ± 3.53 | 133% |
| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | pp1024 | 279.86 ± 45.63 | 100% |
| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | pp2048 | 320.77 ± 5.00 | 124% |
| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | tg128 | 43.51 ± 0.05 | 111% |
| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | tg256 | 43.35 ± 0.09 | 110% |
| qwen2 3B Q4_0_8_8 | 1.69 GiB | 3.09 B | CPU | 64 | tg512 | 42.60 ± 0.31 | 105% |
Q4_0_8_8 offers a nice bump to prompt processing and a small bump to text generation
</details>
</details>
## Which file should I choose?
<details>
<summary>Click here for details</summary>
A great write up with charts showing various performances is provided by Artefact2 [here](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9)
The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.
If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.
If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.
Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.
If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M.
If you want to get more into the weeds, you can check out this extremely useful feature chart:
[llama.cpp feature matrix](https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix)
But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size.
These I-quants can also be used on CPU, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.
</details>
## Credits
Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset.
Thank you ZeroWw for the inspiration to experiment with embed/output.
Thank you to LM Studio for sponsoring my work.
Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski