qwenGPU: Qwen3.5-2B model with compiled GPU layers running CUDA in forward passes
Overview
Developer Vixhal releases qwenGPU, an experimental version of Qwen3.5-2B with 45 compiled GPU layers placed before the model's 24 original layers, under the MIT license on GitHub.
The layers run CUDA compiled to NVIDIA PTX inside the forward pass. According to the developer, every launch was exact and text drift measured KL of about 7e-4. The developer reports speeds of about 10 to 15 tokens per second on Apple MPS. These figures are self-reported.
Written by AI from the articles below · updated Nov 2, 4:56 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- MarkTechPostNewsqwenGPU compiles a GPU into Qwen3.5-2B's weights to run CUDA in forward passes
AIDeveloper Vixhal released qwenGPU, an experimental Qwen3.5-2B model with 45 compiled GPU layers placed before its 24 original layers, under the MIT license on GitHub. The model runs CUDA compiled to NVIDIA PTX inside its forward passes, and the developer reports every launch was exact with a text drift KL of about 7e-4. Speed is about 10 to 15 tokens per second on Apple MPS, and the figures are self-reported.
Heat Trend
- Current comparable heat
- 7
- Peak comparable heat
- 9Nov 2, 5:00 AM ET
- Change in past 24 hours
- –
The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.