Hugging Face Models Scraper
Pricing
from $0.30 / 1,000 model records
Hugging Face Models Scraper
Export public Hugging Face model metadata, downloads, likes, tags, licenses and update dates for model research and organization comparisons.
Hugging Face Models Scraper
Pricing
from $0.30 / 1,000 model records
Export public Hugging Face model metadata, downloads, likes, tags, licenses and update dates for model research and organization comparisons.
Search the public model catalog by model name or ID, for example llama or bert. Leave empty to browse models using the other supported filters. The form prefill is an example, not an API default.
Choose descending catalog order before the result limit: most downloads, most likes, highest trending score, or most recently modified. Omitted or unsupported sort values use downloads in the current runtime.
Save at most 1–1,000 model records. The form and platform schema default to 20; the runtime uses 50 if it receives no limit. Set an explicit value for repeatable calls. A result or time limit can end collection before all matches are returned.
Request extra public metadata for each model, such as structured cardData and repository file counts, when available. This takes longer and does not download files or full card text. If a detail lookup fails, summary data may still be saved with detailFetched=false and a detailError note.
Accepted language parameter, for example en. In the current integration this parameter does not reliably restrict catalog results. Inspect returned tags/languages and filter the exported rows if needed; do not rely on this field for language-only selection.
Optional model task such as text-generation, image-classification, sentence-similarity, automatic-speech-recognition, or text-to-image.
Accepted library parameter, for example transformers. In the current integration this parameter does not reliably restrict catalog results. Inspect returned libraryName and filter the exported rows if needed.
Accepted license parameter, for example apache-2.0. In the current integration this parameter does not reliably restrict catalog results. Inspect returned license/tags and the original model license; do not rely on this field for license-only selection.
Optional inference-provider value sent to the public catalog. Use a valid Hugging Face provider identifier; unsupported values can cause the catalog request to fail. This does not invoke inference or return provider pricing.
35–270 second collection budget, default 270. Collection reserves time for saving results, so it may stop early and return fewer rows. Completed rows are preserved; this is not a guarantee that every requested model will be collected.