Qwenputer: a full RISC-V CPU compiled into Qwen 3.5-4B weights, even running another LLM
CatAstro_Piyush · x · 2026-10-31
A developer demoed nesting compute inside model weights:
- Qwenputer: a full RISC-V CPU compiled into Qwen 3.5-4B's weights, using the model's own Gated DeltaNet layers as RAM. It runs C and MicroPython during forward passes with no external runtime.
- QwenPU: Karpathy's stories260k LLM running on an NPU compiled inside the same model — an LLM inside an LLM.
On DeepMind math: base model scores 55–90%, base + Python tool 95–100%, and Qwenputer hits 97.5–100%.
Related event: Qwenputer: A RISC-V CPU Built Inside Qwen Model Weights(4 posts)→
More from Fun
- Redditor recreates a viral K-beauty video with AI — can you spot the real one? — Zero316 · 2026-11-01
- Bots booking courts at midnight is the new edge: gone by breakfast — chrisfleck · 2026-11-01
- Engineer: AI PCB design doubled my workload — fixing its errors slower than doing it myself — cneuralnetwork · 2026-11-01
- Claude Code drives an oscilloscope over the network to automate waveform debugging — cneuralnetwork · 2026-11-01
- Hacker Turns the WebAI Badge Into a Chess Puzzle Device — jason_mayes · 2026-11-01
- ChatGPT's eerie reading of Lockheed Martin's ominous 'Skunk' teaser goes viral — VoidStateKate · 2026-11-01