VRAM Fit Check Help

See if a model fits your GPU before you download it.

Contents

• Getting started
• Features
• Privacy and permissions
• Troubleshooting
• Getting help

Getting started

1. Install VRAM Fit Check from the Chrome Web Store.
2. Pin it: click the puzzle-piece icon in Chrome's toolbar, then the pin next to VRAM Fit Check.
3. Click the VRAM Fit Check icon to open it.

Features

GPU fit badges

Set your GPU and VRAM once in the extension popup. On any Hugging Face model page or Ollama library model page, each quant/file now shows a fit badge computed from the model's parameter count and the quant's bits-per-weight.

Estimated tokens/sec

The estimate uses a lookup table of typical throughput per GPU tier and model size; it is a rough guide, not a guarantee.

One-click ollama pull tag

Reads the model and tag name directly from the page you're already on and builds the matching pull command; nothing is downloaded or run by the extension.

GPU and VRAM presets

The popup tries to detect and pre-select your GPU automatically on first use (best-effort; browsers don't expose VRAM size, so you always confirm the amount yourself). Open the popup any time to change your GPU or VRAM amount; badges on open model pages update automatically.

On/off toggle

Open the popup and uncheck "Show fit badges on model pages", then reload the model-page tab; badges disappear. Check it again and reload to bring them back.

Privacy and permissions

VRAM Fit Check keeps everything on your device and never sends it anywhere. Permissions it asks for:

• storage: Stores the user's settings and data locally on their device with chrome.storage. Nothing is transmitted.
• clipboardWrite: Copies text to the clipboard when the user presses a copy button.

Troubleshooting

• Nothing happens when I click the icon: open chrome://extensions, make sure VRAM Fit Check is enabled, then reload the page you were on.
• It stopped working after a Chrome update: update the extension at chrome://extensions (Developer mode off, then "Update").
• Does it work in Incognito? No, by design.
• Does this download or run any model? No. It only reads the model size and quant info already shown on the page and computes an estimate locally. Nothing is downloaded, installed or run by the extension.
• How accurate is the VRAM estimate? It's a close approximation based on parameter count and quantization bits-per-weight, plus typical runtime overhead. Actual usage varies by context length, batch size and backend, so treat it as a guide, not an exact number.
• Does this send any data anywhere? No. Your GPU/VRAM choice is stored only in your browser's local storage and never leaves your device.
• Can it detect my GPU and VRAM automatically? It can usually detect which GPU you have (via WebGPU) and pre-select it for you. No browser exposes actual VRAM size to web pages, by design, so you always confirm or enter that amount yourself.

Getting help

Report a bug or ask a question: https://github.com/pinpixels/canirun/issues or pinpixels@gmail.com. Please do not include private data.