Researchers have developed Qwenputer, an experimental model that embeds a functional RISC-V CPU directly within the weights of the Qwen 3.5 language model. This novel approach utilizes the model's own Gated DeltaNet attention layers as RAM, enabling it to perform arithmetic operations and execute code written in C and MicroPython without external tools. The CPU operates entirely within the model's forward passes, offering a unique method for in-model computation and memory. AI
IMPACT Demonstrates novel in-model computation and memory capabilities, potentially influencing future AI architecture research.
RANK_REASON Research paper detailing a novel method for embedding a CPU within an LLM's weights.
Read on Mastodon — mastodon.social →
- Gated DeltaNet
- GitHub
- Hugging Face
- Imanol Schlag
- Jürgen Schmidhuber
- Kazuki Irie
- MicroPython
- ModelScope
- Qwen 3.5
- Qwen-Next
- Qwenputer
- RISC-V
- Schlag et al.
AI-generated summary · Google Gemini · from 4 sources. How we write summaries →