Box v3.1.0
🎉 This is the FIRST Box build where on-device NPU acceleration actually works on Snapdragon and MediaTek phones.
Every previous build shipped the NPU models but crashed on load on those devices. As of 3.1.0, NPU chat and benchmarking genuinely run on Snapdragon and MediaTek hardware. (Pixel / Tensor G5 was already working and is unchanged.)
Plus a smoother chat experience on small NPU models and a refreshed loading screen.
NPU now works on Snapdragon & MediaTek (first time ever)
- First Box release to actually run on the Snapdragon (Qualcomm Hexagon) and MediaTek NPUs — previous versions hard-crashed on load.
- Rebuilt the Qualcomm and MediaTek NPU dispatch libraries to match the LiteRT runtime (2.1.5) and updated the Qualcomm AI stack (QNN 2.47). NPU chat and benchmarking now run instead of crashing. (#81, #83, #88)
- Per-vendor builds so each phone loads the correct NPU driver. (Tensor G5 unchanged.)
Smoother NPU chat on small models
- Long conversations on the Gemma 3 1B NPU model no longer abruptly stop or error when the context fills — Box now slides the context window so the chat keeps going.
- Added safeguards so the small NPU model doesn't get stuck repeating itself or return empty replies. (Snapdragon/MediaTek NPU only — Tensor G5 and GPU/CPU are untouched.)
- Fixed a crash when benchmarking NPU models. (#81)
Polish
- New animated "Initializing model" screen while a model loads.
- Removed the "Experimental" tag from Mobile Actions.
- Tidied up model descriptions.
Thanks to several community members who shared their logs, we have been able to work on getting the NPU running correctly. A special thank you to Manu99it, jnk333, Slayerx96, and soulafein83 for providing essential NPU debugging and logs for Snapdragon and MediaTek devices, Huge thanks to all community members who are actively and consistently reporting bugs it really helps move Box foward ! Thank you :)