What are all the files that are being downloaded?

A common reason for extra storage consumption is caching.

However, with major models like openai/gpt-oss-20b, it’s common for separate files for multiple platforms to be stored within the repository. Downloading the entire repository means all of these get downloaded…

hf download openai/gpt-oss-20b downloads a full snapshot of the repo (i.e., every file in the model repo), not “just one set of weights”. For openai/gpt-oss-20b, the repo contains multiple full-weight artifacts (hence ~41.3GB total).

Files in openai/gpt-oss-20b

Repo root

• .gitattributes
• LICENSE
• README.md
• USAGE_POLICY
• chat_template.jinja
• config.json
• generation_config.json
• model-00000-of-00002.safetensors
• model-00001-of-00002.safetensors
• model-00002-of-00002.safetensors
• model.safetensors.index.json
• special_tokens_map.json
• tokenizer.json
• tokenizer_config.json (Hugging Face)

metal/

• metal/model.bin (13.8GB) (Hugging Face)

original/

• original/config.json
• original/dtypes.json
• original/model.safetensors (13.8GB) (Hugging Face)

Why this becomes ~40–50GB

This repo includes three large “model-weight” payloads:

• Sharded safetensors in the root (model-0000*-of-*.safetensors) totaling ~13.8GB (Hugging Face)
• A single-file safetensors copy under original/model.safetensors (~13.8GB) (Hugging Face)
• A precompiled Metal binary under metal/model.bin (~13.8GB) intended for Apple Metal runtimes (Hugging Face)

That’s already 41.4GB before small metadata/tokenizer files, which matches the repo size shown on the “Files” tab (41.3GB). (Hugging Face)

If you want to avoid downloading everything

Use --include/--exclude patterns. (Hugging Face)

Examples:

• Download only the “original” weights (minimal set recommended in OpenAI’s gpt-oss repo docs):

  hf download openai/gpt-oss-20b --include "original/*" --local-dir gpt-oss-20b/
  

(GitHub)

• Download everything except the Metal and original copies (keep only the root sharded safetensors + configs/tokenizer):

  hf download openai/gpt-oss-20b --exclude "metal/*" --exclude "original/*"
  

(Hugging Face)