Spark-X2.6-1.7B WebGPU

Spark-X2.6-1.7B WebGPU

Runs the Spark-X2.6-1.7B chat model inside your browser on WebGPU, so what you type stays on your device. It suits people with desktop Chrome or Edge who want a private local chat.

Made by mohdmahmodi on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Spark-X2.6-1.7B WebGPU is a static web page that lets you chat with the Spark-X2.6-1.7B model. The model runs on your own graphics card through WebGPU, so nothing you type goes to a server. The page loads the model files on your first visit and the browser keeps them afterwards.

Who it suits: People who want to try a small language model locally in a browser, without an account or a backend.

Industry | Who uses it | How they use it | Example
Software | Web developers | Study how a model runs in a page through WebGPU | Reading the page's lib and app folders
Research | ML hobbyists | Try a 1.7B model on their own graphics card | Asking a few test questions
Privacy work | Cautious users | Chat without sending text to a service | A private draft or note
Education | Teachers | Show students a model running locally | A classroom demo on a desktop

If your machine has no WebGPU browser or spare graphics memory, you can skip it.

Where to use it:

• Desktop Chrome or Edge, on the Space's page at Hugging Face
• The same browsers, opening index.html from a downloaded copy on Windows, macOS or Linux

Use case | Who | What they get back
Private chat | Anyone with a suitable GPU | Replies generated on their own device
Speed check | Developers | A benchmarks page to compare runs
Local demo | Teachers and presenters | A model that works with no server

How to set it up:

• Desktop Chrome or Edge
• A graphics card with roughly 1.5 GB to spare
• Room for a 1.21 GB model download, which the browser caches
• Python only if you want to download the files to your computer

1. Install the Hugging Face hub package.
2. Download the Space with the hf command.
3. Move into the downloaded folder.
4. Open index.html in a web browser.

pip install -U huggingface_hub
hf download mohdmahmodi/Spark-X2.6-1.7B-WebGPU --repo-type space --local-dir Spark-X2.6-1.7B-WebGPU
cd Spark-X2.6-1.7B-WebGPU

How to use it: Open the Space page, or index.html from your copy, and wait for the model to load on the first visit. Then type a message and read the reply. A sidebar link leads to the benchmarks page.

Tool | What it does
index.html | The chat page that runs the model in your browser
benchmarks.html | A separate page of benchmark results, linked in the sidebar

Example: You open the page in desktop Chrome, wait for the one-time model download and type a short question. The reply is generated on your graphics card and shown in the chat.

For newcomers and experts: Open the Space page in desktop Chrome or Edge and send one short message. The first load takes a while because the 1.21 GB model is downloaded once and then cached.

The page names two models. One is the original Spark-X2.6-1.7B and the other is an ONNX version built for WebGPU, which the page fetches from its own model repository. The trade-off is a one-time download and a graphics card with roughly 1.5 GB to spare.

Browser opens the static page
Model files download once and are cached
WebGPU runs the model on your GPU
Replies appear in the chat

Good to know:

• Needs desktop Chrome or Edge and a graphics card with roughly 1.5 GB to spare
• Built as static HTML, so there is nothing to install to use it online
• License: apache-2
• Cost to run: none named
• Data it reaches: it downloads the model from Hugging Face on first use, and your typed text stays on your device
• Not for: browsers other than desktop Chrome or Edge

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: static
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/mohdmahmodi/Spark-X2.6-1.7B-WebGPU
• Files: https://huggingface.co/spaces/mohdmahmodi/Spark-X2.6-1.7B-WebGPU/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.