From f8871238d9d17081ca1eb6731ed259496a168d81 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Victor=20Mu=C5=A1tar?= Date: Tue, 2 Jun 2026 10:57:18 +0200 Subject: [PATCH 1/4] Add Hardware settings doc page Document the dedicated Hardware settings page: how to register the GPUs, CPUs and Apple Silicon you own, and how that drives model compatibility estimates, the profile TFLOPS badge, and the community hardware page. Adds docs/hub/hardware.md and registers it in the Models nav next to Local Apps. --- docs/hub/_toctree.yml | 2 ++ docs/hub/hardware.md | 34 ++++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+) create mode 100644 docs/hub/hardware.md diff --git a/docs/hub/_toctree.yml b/docs/hub/_toctree.yml index 135e60cec1..c883b3a900 100644 --- a/docs/hub/_toctree.yml +++ b/docs/hub/_toctree.yml @@ -183,6 +183,8 @@ title: Model Release Checklist - local: local-apps title: Local Apps + - local: hardware + title: Hardware - local: models-faq title: Frequently Asked Questions sections: diff --git a/docs/hub/hardware.md b/docs/hub/hardware.md new file mode 100644 index 0000000000..3201b82f80 --- /dev/null +++ b/docs/hub/hardware.md @@ -0,0 +1,34 @@ +# Hardware + +Tell the Hub which compute hardware you own (GPUs, CPUs, or Apple Silicon) and it will help you find models that run on your setup, show your setup on your profile, and let you compare it with the rest of the community. + +Manage it any time from your [Hardware settings](https://huggingface.co/settings/hardware). + +## Add your hardware + +On your [Hardware settings](https://huggingface.co/settings/hardware) page, add each piece of hardware you own: + +1. Pick a **type**: GPU, CPU, or Apple Silicon. +2. Choose the **provider** and **model** (for example, NVIDIA RTX 4090). +3. Set the **memory** (VRAM, RAM, or unified memory) and the **number of units** you have. + +Add as many items as you like. If you own several, mark one as your **primary** hardware so it appears first on model pages. + +> [!TIP] +> Your hardware is private by default. Use the **Publicly Visible** toggle to show it on your profile. + +## See which models fit your hardware + +On model pages that offer [GGUF](./gguf) or MLX files, a **Hardware compatibility** panel estimates whether each quantization will run on your saved hardware, so you can choose a size that fits before downloading. Pair it with [Local Apps](./local-apps) to go from "will it run?" to running it in a couple of clicks. + +## Share and compare + +Once your hardware is public: + +- A **TFLOPS** badge appears on your profile, summarizing your total compute power. +- You can browse [what the community is running](https://huggingface.co/hardware) and see how your setup compares across GPUs, CPUs, and Apple Silicon. + +## Next steps + +- [Use AI Models Locally](./local-apps) — run models with your favorite local app. +- [Local Agents with llama.cpp](./agents-local) — build a coding agent on your own hardware. From 47067f03fbb18b8c31b02a03acd2d2679fddff59 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Victor=20Mu=C5=A1tar?= Date: Tue, 2 Jun 2026 11:04:13 +0200 Subject: [PATCH 2/4] Put Hardware before Local Apps in the nav --- docs/hub/_toctree.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/hub/_toctree.yml b/docs/hub/_toctree.yml index c883b3a900..870ce08c51 100644 --- a/docs/hub/_toctree.yml +++ b/docs/hub/_toctree.yml @@ -181,10 +181,10 @@ title: Models Download Stats - local: model-release-checklist title: Model Release Checklist - - local: local-apps - title: Local Apps - local: hardware title: Hardware + - local: local-apps + title: Local Apps - local: models-faq title: Frequently Asked Questions sections: From e29885dd112fe6e2807b72f61e21192da03fc8e5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Victor=20Mu=C5=A1tar?= Date: Tue, 2 Jun 2026 13:44:53 +0200 Subject: [PATCH 3/4] Address review: hardware is public by default, wording tweaks - Fix the visibility default: hardware is public by default; turn off Publicly Visible to keep it private (per review). - Reframe the intro so profile sharing and the privacy control are clear. - Minor wording: 'get up and running', 'estimated total compute power'. --- docs/hub/hardware.md | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/docs/hub/hardware.md b/docs/hub/hardware.md index 3201b82f80..e5b37103c8 100644 --- a/docs/hub/hardware.md +++ b/docs/hub/hardware.md @@ -1,6 +1,6 @@ # Hardware -Tell the Hub which compute hardware you own (GPUs, CPUs, or Apple Silicon) and it will help you find models that run on your setup, show your setup on your profile, and let you compare it with the rest of the community. +Tell the Hub which compute hardware you own (GPUs, CPUs, or Apple Silicon) and it will help you find models that run on your setup. It also appears on your public profile, so you can compare your setup with the community, and you can make it private any time. Manage it any time from your [Hardware settings](https://huggingface.co/settings/hardware). @@ -15,17 +15,17 @@ On your [Hardware settings](https://huggingface.co/settings/hardware) page, add Add as many items as you like. If you own several, mark one as your **primary** hardware so it appears first on model pages. > [!TIP] -> Your hardware is private by default. Use the **Publicly Visible** toggle to show it on your profile. +> Your hardware is public by default. Turn off the **Publicly Visible** toggle to keep it private. ## See which models fit your hardware -On model pages that offer [GGUF](./gguf) or MLX files, a **Hardware compatibility** panel estimates whether each quantization will run on your saved hardware, so you can choose a size that fits before downloading. Pair it with [Local Apps](./local-apps) to go from "will it run?" to running it in a couple of clicks. +On model pages that offer [GGUF](./gguf) or MLX files, a **Hardware compatibility** panel estimates whether each quantization will run on your saved hardware, so you can choose a size that fits before downloading. Pair it with [Local Apps](./local-apps) to get up and running in a couple of clicks. ## Share and compare -Once your hardware is public: +While your hardware is public: -- A **TFLOPS** badge appears on your profile, summarizing your total compute power. +- A **TFLOPS** badge appears on your profile, summarizing your estimated total compute power. - You can browse [what the community is running](https://huggingface.co/hardware) and see how your setup compares across GPUs, CPUs, and Apple Silicon. ## Next steps From e85ccf777bbaf930f52278279ea6290eb3f50266 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Victor=20Mu=C5=A1tar?= Date: Mon, 8 Jun 2026 11:22:29 +0200 Subject: [PATCH 4/4] Add Hardware settings screenshot (light/dark) --- docs/hub/hardware.md | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/docs/hub/hardware.md b/docs/hub/hardware.md index e5b37103c8..95307aa91c 100644 --- a/docs/hub/hardware.md +++ b/docs/hub/hardware.md @@ -4,6 +4,11 @@ Tell the Hub which compute hardware you own (GPUs, CPUs, or Apple Silicon) and i Manage it any time from your [Hardware settings](https://huggingface.co/settings/hardware). +
+Hardware settings page in light mode. + +
+ ## Add your hardware On your [Hardware settings](https://huggingface.co/settings/hardware) page, add each piece of hardware you own: