divagr18/QwenPU
Model
Can I trust it?
Be careful with this one.
Warnings Check
Very new: it was created yesterday. Wait a little before installing it.
Stars from real accounts No data
We haven't checked its stars for fakes yet.
Active OK
Last change today.
Licence Careful
No licence: legally you can't reuse the code, unless the README says otherwise.
Quick checks from public data. They help, but they are no guarantee.
What it is
A Qwen3.5-4B model modified with a 32-bit RISC-V CPU and a vector unit compiled into its weights, which runs Karpathy's stories260K Llama as a C program inside the model's forward passes. It is a research demonstration of in-weights computation, not a production model, and needs…
Who it's for
Researchers and developers interested in mechanistic interpretability and in-weights computation in language models
How to use it
Install
uv sync --extra hfFrom its README. Check it there before running it.
Ask your AI to set it up
Paste this into Claude, ChatGPT or your coding assistant.
Help me install and set up divagr18/QwenPU in my project. Follow its README: https://github.com/divagr18/QwenPU. Explain each step and check it is safe before running anything.
Star curve
Get alerts for repos like this
We'll email you when something in Models & inference starts rising. Free, once a week, only what matches.
23,609 GitHub repos watched and ranked every hour. Figures as of 1 November 2026.
What arrives: one email on Mondays with the repos that match, never more. See last week's. Privacy
For maintainers
Add this badge to your README. It shows the repo's place on the radar, updates every hour and links back to this page.
[](https://repoalerts.com/repo/divagr18/QwenPU?ref=badge)Similar repos
- divagr18/qwenputer
An experimental model: Qwen3.5-4B with a 32-bit RISC-V CPU compiled into its weights, using Gated DeltaNet state as RAM. Qwen calls a calc tool, and the compiled layers run it on the CPU inside the fo
- MaxDevv/lean-qwenflash
Inference runtime for the Qwen3.8-Flash-Next 125B MoE model that streams uncached expert weights from NVMe via O_DIRECT into a bounded GPU cache, targeting full 262k context on 16 GB VRAM and 8 GB sys
- dbreunig/nesformer
Runs an NES 6502 CPU, its work RAM and a cartridge's program ROM inside the weights of a stock Qwen3.5 checkpoint, with weights compiled rather than trained. A small host program handles drawing, cont
- lyramakesmusic/neuron-explorer
A steering-first explorer for Qwen3.5-4B that pushes every MLP neuron in a layer up and down on a few prompts, keeps all rollouts, and lays the results out in a 3D view of the model. It also runs a li
- freriadoc/Avalon
A fork of llama.cpp tuned for running the Qwen3.8-Flash-Next 125B MoE model on a single consumer AMD GPU with experts offloaded to system RAM. It uses an expert cache in VRAM, CPU computation of missi
- prabanta-dev/janas
A C inference engine for running large mixture-of-experts language models on ordinary computers, with or without a GPU. Weights stay on disk, and the experts each token needs are streamed from an NVMe