pd95/ gptoss-mlx:20b-mxfp4

88 1 week ago

Experimental MLX-backed GPT-OSS 20B for custom Ollama builds. Requires GptOssForCausalLM support and is not compatible with the regular Ollama app.

tools thinking 20b 
ollama run pd95/gptoss-mlx:20b-mxfp4

Details

1 week ago

50d14eb2f9db · 12GB ·

json
{ "architectures": [ "GptOssForCausalLM" ], "attention_bias": true, "attention_dropout": 0.0, "eos_t
json
{ "bos_token_id": 199998, "do_sample": true, "eos_token_id": [ 200002, 199999, 200012 ], "pad_token_
json
{ "bos_token": "<|startoftext|>", "eos_token": "<|return|>", "pad_token": "<|endoftext|>" }
json
{ "version": "1.0", "truncation": null, "padding": null, "added_tokens": [ { "id": 199998, "content"
json
{ "added_tokens_decoder": { "199998": { "content": "<|startoftext|>", "lstrip": false, "normalized":
license
Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR US
params
{ "temperature": 1 }
template
<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI. Knowledge cutof
339 tensors

Readme

gptoss-mlx

Experimental MLX-backed Ollama build of GPT-OSS 20B.

This model requires a custom Ollama build with experimental MLX safetensors support and GptOssForCausalLM support. It will not run on the regular public Ollama app/build.

Tags

  • pd95/gptoss-mlx:20b - recommended default, currently MXFP4
  • pd95/gptoss-mlx:20b-mxfp4 - explicit MXFP4 artifact

Source

This model is an experimental Ollama/MLX conversion of:

https://huggingface.co/openai/gpt-oss-20b

GPT-OSS was released by OpenAI.

License and Use Policy

GPT-OSS is released under the Apache License 2.0:

https://huggingface.co/openai/gpt-oss-20b/blob/main/LICENSE

Use is also subject to the upstream model terms and safety guidance:

https://huggingface.co/openai/gpt-oss-20b

This repository only republishes a quantized MLX/Ollama artifact. It does not change the upstream GPT-OSS license or usage terms. Please review the upstream Apache-2.0 license and model card before use or redistribution.

Notes

  • Architecture: GPT-OSS 20B
  • Quantization: MXFP4
  • Context window: 128K
  • Capabilities: text, tools, thinking
  • Runtime: experimental Ollama MLX runner

This model is intended for testing the experimental GPT-OSS MLX path in Ollama.