01
VRAM fit, per quant.
Set your card once. Every quantization on the model page is auto-classified fits, tight, or won't fit — with the math shown below.
Features, install instructions, and frequently asked questions. The extension itself is local-only — no account or API key required.
Eight layers of tooling on every Hugging Face page you visit. No login, no tracking, no API key.
Set your card once. Every quantization on the model page is auto-classified fits, tight, or won't fit — with the math shown below.
Apps (vLLM, llama.cpp, Ollama, LM Studio, Jan, MLX LM, Draw Things, TGI, SGLang) and libraries (Transformers, Diffusers, GGUF, MLX, Sentence-Transformers, PEFT, Safetensors) — flagged early on every model and listing card.
Browsing /models? Cards get tagged with size + fit + format + last-updated. Dim or hide non-fits. Tick your runtimes — we redirect to HF's own URL filter (?apps=…&library=…) so the server pre-narrows the results.
Per-quant ↓ on every GGUF row. Whole-repo ↓ for safetensors / AWQ / GPTQ / FP8 / NVFP4 / diffusion checkpoints. Sharded files reassembled into one tidy folder. No CLI, no python, no auth flow.
Land on /<org>? A quick-nav block lets you jump to their models with your runtime filter pre-applied. Datasets, Spaces, Collections, Papers, Posts — all one click, no scrolling.
Toggle "Theme HF page" and Hugging Face goes pink (Aurora), emerald (Forest), brutalist (Mono), or stays cream (Daylight). Hijacks Tailwind's dark mode and retints the dark surfaces with your palette. The site you stare at all day, in your colors.
Hide HF clutter you never use: promo banners, the inference widget, "Use this model", trending carousels, the footer. Or paste your own selectors.
Hover any model link anywhere on Hugging Face — a dossier appears with parameters, license, base model, formats, and a fit verdict. Pin it so it doesn't auto-hide. Drag it anywhere on the page. Hit ↓ download repo without ever clicking through.
Manifest v3 · works on Chrome, Edge, Brave, Arc, Vivaldi. ~40 KB.
Pick from 8 / 12 / 16 / 24 / 48 / 80 — or type a custom value. Tune headroom and KV-cache dtype.
Fit verdicts, runtime badges, and dossiers appear inline. Cards that won't fit dim or disappear.
No. The extension only reads the Hugging Face DOM you're already looking at and computes everything in your browser. There is no analytics, no telemetry, no account, no API key.
Per quant: file size + KV-cache estimate (configurable fp16 / fp8) + headroom for activations. Diffusion repos use a 3 GB minimum headroom regardless. Fits / tight / over thresholds map to colour bands.
It works on text models with quants on the page, on diffusion repos, and on listing pages. If a page genuinely lacks information (e.g. unlabelled custom forks), Hugging Fit says so rather than guessing.
The popup theme drives the floating dossier, listing-card badges, and profile mini-panel — pick once, it travels. Toggle Theme HF page and the same palette retints Hugging Face itself by hijacking its Tailwind dark class and retinting the dark surfaces.
Downloads/huggingface/<org>_<repo>/<file>. Sharded GGUFs (…-00001-of-00005.gguf) all land in the same folder. Anything > 5 GB or > 3 files asks for confirmation first — no surprise 200 GB pulls.
Yes. Tick vLLM, Ollama, llama.cpp, plus tasks / licenses / 4-bit / MoE — turn on Auto-apply to HF URL, and visiting /models redirects to ?apps=…&library=…&pipeline_tag=…&license=…&other=…. HF filters server-side, you scroll less.
Manifest v3 + chrome.storage means Chromium browsers work today. A Firefox port is on the list once a usable WebExtensions polyfill ships.
Five seconds to install. Free. No tracking. Your VRAM thanks you.
Chrome Web Store listing pending review