Developer releases Qwenputer, a Qwen3.5-4B model with an embedded RISC-V CPU
Overview
Developer Divyansh Agrawal has released Qwenputer, an experimental model built on Alibaba's Qwen3.5-4B that embeds a 32-bit RISC-V CPU and memory inside the neural network, so it can run programs and exact calculations without external tools such as Python.
The project reports large-integer multiplication accuracy rising from 55% to 97.5%, and 100% on the NUPA long-arithmetic benchmark versus 48.6% for the base model. The model is open-sourced on Hugging Face and GitHub and runs on a single GPU with 24GB of memory. It grows to about 18.8GB and runs slower than the base model, and it needs a dedicated driver rather than loading like a standard large language model.
Written by AI from the articles below · updated Oct 30, 5:41 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- IThome · AINewsDeveloper builds Qwenputer, embedding a RISC-V CPU inside Qwen3.5-4B
AIDeveloper Divyansh Agrawal released Qwenputer, an experimental model built on Alibaba's Qwen3.5-4B that embeds a 32-bit RISC-V CPU and memory into the network. It raised large-integer multiplication accuracy from 55% to 97.5% and scored 100% on the NUPA long-arithmetic benchmark, versus 48.6% for the base model. The model grows to about 18.8GB and runs slower, and it requires a dedicated driver. It is open-sourced on Hugging Face and GitHub and runs on a single 24GB GPU.
Heat Trend
Not enough continuous observations to show a trend yet.