Radar ·
QwenPU: a small LLM running inside the weights of Qwen 3.5-4B
A briefing describes QwenPU, a stories260k model running on an NPU compiled into the weights of Qwen 3.5-4B, and Qwenputer, a RISC-V CPU embedded in the same weights, with Gated DeltaNet layers used as RAM.
According to the briefing, the author describes QwenPU, a stories260k LLM running on an NPU compiled into the weights of Qwen 3.5-4B. A cited post describes Qwenputer, a RISC-V CPU embedded in the same weights, which uses the Gated DeltaNet layers as RAM. The material shows that computation can be compiled into a model's weights, a topic relevant to readers studying Gated DeltaNet and model internals.
The source is a short summary and gives no details on methodology, performance or reproduction. To assess the claims, consult the author's original post and the cited post, checking dates, code and published results. Rota Nacional does not implement or solve this technique. If you use AI to study the topic, do not paste proprietary code, internal data or documents containing personal data into it: the platform detects CPF, CNPJ, e-mail, phone numbers and names before any model runs and applies your organization's policy, using markers, removal or a block.
Original source: QwenPU: um LLM pequeno rodando dentro dos pesos do Qwen 3.5-4B