Qwenx86 Debugger

Qwenx86 Debugger

Replays recorded runs of Qwenx86, a Qwen model with an x86 computer built into its weights, and shows registers, code, RAM and tool output step by step in your browser.

Made by Quazim0t0 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: Qwenx86 Debugger is a browser page that replays recorded runs of Qwenx86, a Qwen3.5-4B model with a 64-bit x86 computer built into 13 added layers of its weights. You pick a run, press play, and watch the registers, code, RAM and tool output change as the model works.

Who it suits: It suits people who are curious how a language model can run real machine code, and researchers who want to check a recorded run in detail.

Industry | Who uses it | How they use it | Example
AI research | Model researchers | Inspect how a model does machine-level work | Check that every machine token in a run matched the reference machine
Education | Teachers and students | Show how registers and RAM change while a program runs | Watch the C-compiler run print 42
Software | Systems developers | Read the disassembled code at the program counter | Follow the Python tool run on the RISC-V path

Anyone who only wants a chat assistant can skip it, since the page replays recordings and does not take new questions.

Where to use it:

• A web browser, on the Space's page on Hugging Face
• A web browser on Windows, macOS or Linux, from a local copy of the files

Use case | Who | What they get back
Replay the C-compiler run | Anyone curious about the model | A C program written by Qwen, compiled and run inside the weights, printing 42
Inspect machine state | Researchers | Registers, disassembled code and all 64 KB of RAM at each step
Review the memory runs | Developers | Files, stored passages, pictures and exact arithmetic shown from real runs

How to set it up:

• A web browser, which is all the online page needs
• Python with pip, only for a local copy

1. Open the Space's page on Hugging Face to use it online. Nothing needs installing.
2. For a local copy, install the Hugging Face hub package with pip.
3. Download the Space's files into a folder named Qwenx86-Debugger.
4. Move into that folder and open index.html in a web browser.

pip install -U huggingface_hub
hf download Quazim0t0/Qwenx86-Debugger --repo-type space --local-dir Qwenx86-Debugger
cd Qwenx86-Debugger

How to use it: Open the page, choose a run and press play. The top banner replays the C-compiler run. Under it, each run lays out the chat, the registers, the code, the RAM, what the tool printed and a timeline. The model file ships with six built-in tools.

Tool | What it does
c | Compiles and runs C programs
notes | Keeps notes in the machine's RAM
science | Looks up CODATA constants, unit data and elements from read-only data
python_doc | Holds 1,003 compressed API signatures
python | Runs MicroPython
calc | Does exact fractions

Example: Pick the notes run and press play. The first question saves a note and the second asks for it back. You see each tool call, the computer's reply with its machine token count, and the RAM bytes that were just written, lit in amber.

For newcomers and experts: Think of the page as a recording you can scrub through. Start with the banner, press play on the C-compiler run, and watch the program print 42. Then open the notes run to see a saved note come back.

Qwenx86 puts a 64-bit x86 CPU circuit into Qwen3.5-4B's weights, so tool calls run inside the model. The Space shows the reference machine's states for each run: RAX to R15 and RFLAGS, code at the program counter and a timeline of machine tokens grouped in block verify passes. The chip layers checked up to 256 tokens per pass, and every token matched. The Python run uses the core's RISC-V path, so it shows x1 to x31 instead.

Qwen writes a tool call
The computer in the weights runs it
It produces machine tokens
Block verify passes check them
The page replays each machine state

Good to know: Every run is a saved recording from a single 16 GB GPU using the compact checkpoint, so the page never runs the model live.

• Memory: the memory section works alongside the model under Python control, and Qwen's replies do not change with the memory full or empty
• License: MIT
• Cost to run: none named
• Data it reaches: the recorded run data bundled in its own data folder

On Hugging Face, 13 November 2026

• Likes: 7
• Built with: static
• License: mit

Links

• Space: https://huggingface.co/spaces/Quazim0t0/Qwenx86-Debugger
• Files: https://huggingface.co/spaces/Quazim0t0/Qwenx86-Debugger/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.