MXFP4
Also: MXFP4_MOE
A compact 4-bit format some newer models are released in, which keeps their files small.
The fuller definition
MXFP4 is a 4-bit floating-point weight format with shared scaling per small block of weights. Some recent models, including gpt-oss, are released in it.
- Weights
4-bit values
- Blocks
Shared scale
- File
Smaller model
A simplified relationship, not a screenshot or a complete design.
A concrete example
gpt-oss-20b ships as a 12.11 GB MXFP4 GGUF file.
The distinction that matters
MXFP4 is a format the model was trained or released in, not just a compression applied afterwards.
Related terms
Quantization
Storing model numbers with less detail. This can save memory. The effect on quality depends on the model and the task.
Training & generationQuantization labels (Q4_K_M and others)
The codes in model file names, like Q4_K_M. The number is roughly how many bits each setting uses; smaller is lighter and slightly less precise.
Hardware & hostingGGUF
The file format most local AI apps use for downloaded models.
Where it is taught
How models learn from examplesSource and further reading
Hugging Face · LLM courseThe definition and example are written for this library. The link gives background, official documentation or legal authority as labelled. Legal scope depends on the jurisdiction and current text; product features depend on release and settings.