GPT-OSS-20B — Original MXFP4 Reference Weights (Mirror)
⚠️ IMPORTANT
This is an archived mirror and NOT the original upstream repository maintained by OpenAI.
It is not affiliated with, endorsed by, or officially supported by OpenAI.
It contains only the original MXFP4 reference checkpoints of GPT-OSS-20B for archival, research, and conversion purposes.
These weights are NOT inference-ready and cannot be used directly for generation without conversion to an appropriate runtime format.
See upstream for inference-ready usage.
Repository Scope
This mirror stores a minimal set of artifacts required to preserve the GPT-OSS-20B model in its original MXFP4 quantized format.
The repository does not include inference binaries, converted weights, or example usage code.
📦 Contents
📂 Model Weights (Original)
Located in the original/ directory:
*.safetensors
→ Original MXFP4 reference weights
These files are the reference parameters released by OpenAI for GPT-OSS-20B.
They represent the ground-truth weight values in MXFP4 quantization.
📄 Model Configuration & Metadata
config.json
→ Model architecture definition (layers, dimensions, MoE structure)- Dtype / quantization metadata
→ Documents the MXFP4 numeric format used by the weights LICENSE
→ Apache License, Version 2.0
❌ Not Included
To avoid ambiguity, this repository does NOT include:
- Tokenizer files (
tokenizer.json,tokenizer.model, etc.) - Inference-ready weights (Transformers / vLLM / Metal)
- Runtime binaries or compiled artifacts
- Example inference code
Tokenizer and inference artifacts are available from the upstream OpenAI repositories.
About MXFP4 Quantization
GPT-OSS models were released with MXFP4 quantization of their MoE weights, a custom 4-bit floating format designed to reduce memory requirements.
- Allows GPT-OSS-20B to fit within ~16 GB of GPU memory
- Requires specialized tooling and kernels for inference
- This repository preserves the original MXFP4 semantics for future use
MXFP4 weights are not directly usable in most standard inference engines without conversion.
Intended Use Cases
This repository is suitable for:
- Long-term archival and backup
- Research into model parameters
- Converting weights into runtime-specific formats
- Fine-tuning workflows that support MXFP4
- Audit and reproducibility work
It is not intended for direct inference or deployment.
Upstream References
- Model card and download: https://huggingface.co/openai/gpt-oss-20b
- Main GPT-OSS repository (reference code): https://github.com/openai/gpt-oss
- GPT-OSS announcement: https://openai.com/index/introducing-gpt-oss
License
This repository and the included artifacts are distributed under the Apache License 2.0.
All upstream model rights remain with OpenAI under the terms of the Apache 2.0 license.
Cooperation Notice
This mirror is maintained for archival and research purposes in good faith.
If the original rights holder or upstream maintainers request clarification, updates, or removal, the maintainer will cooperate where reasonable and consistent with the Apache License 2.0.
Requests may be raised via repository Discussions or Issues.
- Downloads last month
- 1