Hugging Face co-founder Julien Chaumond has rolled out a hardware filter on the public Models page that surfaces only those models confirmed to run on devices such as the Apple M4 Max, NVIDIA RTX 4090, or various Intel CPUs. The tool draws from user-configured hardware settings and works across vanilla weights plus GGUF and MLX formats, addressing repeated requests for simpler local-inference discovery.
@mervenoyann @alvarobartt @mihai673 @xpasky @huggingface @victormustar it's live!
http://huggingface.co/changelog/filter-models-by-hardware
@alvarobartt @mihai673 @xpasky @huggingface @julien_c @victormustar and MLX, would be nice to have for bare PT models 👀
After a single hardware profile is saved, the same criteria apply across all model searches and can be shared via URL even with logged-out visitors, removing the need to re-enter specs on every visit.
While M4 Max, RTX 4090, and Intel CPUs are the highlighted examples, the full roster of supported SKUs and the matching logic remain unspecified for now.
Many users praised Hugging Face for adding hardware filters to its models page because the feature makes it easy to find local AI models that actually run on their hardware and saves time guessing compatibility.
No Digg Deeper questions have been answered for this story yet.
NEW on Hugging Face: Hardware filters 🖥️
A new Hardware filter on the Models page results to models that fit a specific GPU, CPU, or Apple Silicon chip, so you only see what will actually run on your machine. Set the hardware you want from your Hardware settings.
It stacks with the page's other filters and is shareable via URL, so a link like llama.cpp&hardware=apple-m4-max points anyone, even logged-out visitors, at the llama.cpp models that fit an Apple M4 Max.
A study from @Stanford showed that 71.3% of chatgpt queries could be accurately answered by a local model. I suspect a major part of enterprise AI workloads could be run locally too for free (compared to the massive costs of frontier API cost).
Also, it reduces the risk of these workloads being taken away from you because you own the models instead of renting them - which sounds like a good idea these days haha.
That's why we're introducing the ability for everyone to filter AI models on @huggingface based on your local hardware.
For me, there are 800k+ public models that fit on my M5 24GB and that I can use easily thanks to llamacpp.
Let's go local AI!
waited for this feature so long
you can now filter models for your hardware 🔥 and it works not only for GGUF/MLX compatible models but vanilla weights too 😍
@mervenoyann OH YES!!! I was waiting for this one too! Hugging Face adding pieces to the Local AI puzzle week after week 💪
you can now filter huggingface models capable to run on your hardware
@ClementDelangue @Stanford So happy to see this feature. Thank you @ClementDelangue !! 🚀
@victormustar i tried it. it told me i'm poor and go play video games
thanks i guess
@petrroyce what do you have?
get started at http://hf.co/models
@ClementDelangue @Stanford I predict that 99.99999% of current and future AI use cases can be handled by a <32b model. Smaller when we make the models specific rather than generalized.
@ClementDelangue @Stanford Not any of my own queries lol
@mervenoyann Woohoo. I appreciate this must have taken a fair bit of work behind the scenes to get right.
@ClementDelangue @Stanford Sweet! An API to query compatible models based on hardware config would be super valuable.
on top of this, you don't need to put your own hardware in settings to filter for hardware
more info here 🫶🏻 https://huggingface.co/changelog/filter-models-by-hardware
waited for this feature so long
you can now filter models for your hardware 🔥 and it works not only for GGUF/MLX compatible models but vanilla weights too 😍
@victormustar 🔥🔥
@briancaffey @Stanford lots of users requested it over the years!
@victormustar YES 🙌
@mervenoyann this and that HF cli command that pops up when you open the .safetensors file 👌
@victormustar M4 and M3 Max
on the bright side, I still can do models, GPU poor cannot stop me
@ClementDelangue @Stanford "You own the models instead of renting them" is the same logic behind self-hosting anything- calendars, contacts, AI, doesn't matter.
Once your data or your tools live on someone else's infrastructure, you're one policy change away from losing access to your own workflow.