Qwen-Image 2.1 Turbo (20B) runs fully on a phone NPU: 40s per image on Snapdragon 8 Gen 1
shirosoro · reddit · 2026-11-09
An indie dev ported Qwen-Image 2.1 Turbo (20B) to run entirely on the Hexagon NPU of older Snapdragon phones (8 Gen 1 / 8+ Gen 1) via the Android app HexAnima. The full pipeline runs on the NPU: the 20B DiT in 4 steps, the Qwen3-VL-8B text encoder, and the VAE, supporting both text-to-image and photo editing.
Real-world numbers on a 12 GB phone: 40-50 s per generated image, 1 min per edit. Weights are streamed from storage, keeping RAM use around 1.2 GB, though the download is 14 GB.
Limits: only tested on a 12 GB device, requires app v1.4.0+; weights are under the non-commercial Qwen Research License. Code is MIT open source with sample outputs and prompts at github.com/ayanoby/hexanima
More from Infra
- Getting Omarchy running on Snapdragon Surface Pro: custom kernel plus SSH tunnel for OpenClaw — vincent_koc · 2026-11-09
- Ex-PlayStation Boss: A $300 PS5, Not a $1,200 PS6, Would Be True Innovation — aakashgupta · 2026-11-09
- Rumor: Megaspeed's SEA datacenter shell route triggered US crackdown, B300 spot prices soar — basedjensen · 2026-11-09
- NVIDIA's AutoCompress Beats Minitron by Up to 6.57 Points With Hardware-Guided LLM Pruning — nvidia · 2026-11-09
- llama-modchip Runs DeepSeek's 284B Model Entirely in 64GB RAM With Zero GPUs — solyarisoftware · 2026-11-09
- China sets fiber record: 64 Tbps each way over 1,007 km of hollow-core fiber — AIFlow_ML · 2026-11-09