Open Source

Hugging Face adds 'Base only' toggle to filter out finetunes and quantized models

No more scrolling through thousands of LoRA adapters to find the original model.

Deep Dive

A feature that was requested a lot: https://huggingface.co/models?base_model_relation=base

Key Points
  • New 'Base only' filter lives at huggingface.co/models?base_model_relation=base
  • Hides all fine-tuned, quantized, and adapter variants from search results
  • Addresses a long-standing community request to reduce model page clutter
  • Helps developers quickly identify original foundation models like Llama 3 or Mistral

Why It Matters

Saves ML engineers time by showing only original base models, reducing search noise on the largest open model hub.