# qwenGPU compiles a GPU into Qwen3.5-2B weights

Digest AI · Generative AI & Models · published 2026-11-02T07:28:29Z

Canonical: https://digestai.news/story/qwengpu-compiles-a-gpu-into-qwen3-5-2b-weights

## Summary

qwenGPU is an experimental model that embeds a 4‑lane SIMT GPU inside the weights of Qwen3.5‑2B. Developed by Vixhal and released on GitHub under MIT, it compiles CUDA to PTX and runs inside the model’s forward passes.

The model adds 45 untrained GPU layers to the 24 layers of Qwen3.5‑2B, for a total of 69 layers. It runs on Apple silicon, NVIDIA GPUs or CPU with about 16 GB free memory. Performance tests show exact launches, a KL divergence of about 7e‑4 from the base model, and a reference interpreter that matches an NVIDIA Tesla T4 on 49 of 49 kernels. Throughput is 10‑15 tokens per second on Apple MPS, indicating a research demo rather than production use.

qwenGPU uses Qwenputer’s transformer compiler and a Gated DeltaNet state to act as registers, ROM, and global memory. The demo includes a dot‑product kernel that returns 32 in 47 tokens. The project is still experimental, with no prebuilt checkpoint and limited reliability for a 2B model.

## Key points

- qwenGPU adds 45 GPU layers to Qwen3.5‑2B, totaling 69 layers.
- Runs on Apple silicon, NVIDIA GPU or CPU with ~16 GB free memory, 10‑15 tps on Apple MPS.
- Reference interpreter matches Tesla T4 on 49/49 kernels, KL divergence ~7e‑4.

## Why it matters

qwenGPU shows a GPU can be embedded inside a language model’s weights, enabling exact CUDA execution during inference. This could reduce overhead and inspire hybrid compute models, but current throughput limits it to research rather than production.

## Sources

1. [Meet qwenGPU: Qwen3.5-2B With 45 Compiled GPU Layers That Run CUDA](https://marktechpost.com/2026/11/01/qwengpu-simt-gpu-compiled-into-qwen3-5-2b-weights-runs-cuda) (MarkTechPost, 2026-11-02)

Part of the developing story: [Qwen3.8 Flash Next Optimization Sprint](https://digestai.news/thread/inco-ai-releases-dflash-2-draft-model-for-qwen3-8-flash-next) (5 stories)

## Cite

Digest AI, "qwenGPU compiles a GPU into Qwen3.5-2B weights", 2 November 2026, https://digestai.news/story/qwengpu-compiles-a-gpu-into-qwen3-5-2b-weights

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/qwengpu-compiles-a-gpu-into-qwen3-5-2b-weights.json