Hugging Face
Platform for hosting, sharing and discovering machine learning models, datasets and demos.
Valued as the default place to publish and find open models and datasets123, with simple distribution and versioning for large files45, and as a way to compare inference providers for open models.6 Commenters raise no specific complaints; it suits people downloading open-weights models, publishing datasets, or testing models through hosted inference.789
Why people recommend it
Model publishing hub · 3
Commenters see it as the de facto place where projects and open-weights models get published.
“it's basically become a defacto "publish your project here" thing.”
“Huggingface.co is where many of them are published.”
Show 1 more
“if not for arxiv and huggingface, the field of ML would be nowhere near where it is today”
Dataset hosting · 2
Used to find datasets and to publish them, with distribution and versioning already handled.
“you can pick some agent trace dataset on https://huggingface.co/datasets?format=format%3Aagent-traces”
“I went with Hugging Face mainly because it is simple and already handles distribution and versioning.”
Inference provider comparison · 1
Lets you pick a model, see which inference providers serve it, and access it directly.
“One good way to find inference providers for open models is through hugging face”
Large file hosting · 1
Hosts files too large for GitHub.
“it's basically become a defacto "publish your project here" thing.”
Everyday tool · 2
Some commenters use it all the time and credit it with advancing the ML field.
“I use huggingface all the time.”
“if not for arxiv and huggingface, the field of ML would be nowhere near where it is today”
Caveats and dissent
No one in these comments warned against it.
Who recommends it in their own words
Threads where it came up
- Nvidia agrees to acquire Hugging Face for $13B 2 mentions
- ArXiv receives multiyear commitments to support it as an independent nonprofit 1 mention
- DeepSeek V4 Flash on a Single AMD MI300X 1 mention
- Ask HN: What are you using for LLM inference in production? 1 mention
- The Unbearable Cheapness of Open Weight Models 1 mention
- Show HN: Hacker News archive (47M+ items, 11.6GB) as Parquet, updated every 5m 1 mention
Data: hugging-face.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 7
“if not for arxiv and huggingface, the field of ML would be nowhere near where it is today”
“it's basically become a defacto "publish your project here" thing.”
“I use huggingface all the time.”
“you can pick some agent trace dataset on https://huggingface.co/datasets?format=format%3Aagent-traces”
“One good way to find inference providers for open models is through hugging face”
“Huggingface.co is where many of them are published.”
“I went with Hugging Face mainly because it is simple and already handles distribution and versioning.”