NPU
31 updates on NPU.
Quant.npu runs 4-bit LLMs up to 15.1 percent faster on a Qualcomm NPU
Eight authors report a fully static integer quantisation method that runs 4-bit models up to 15.1 percent faster on a Qualcomm SM8650 NPU.
iPhone 16 Pro settles 41.5 percent below peak over 20 back-to-back prompts
Four authors ran Qwen 2.5 1.5B 20 times in a row on four edge platforms and report the iPhone 16 Pro settling 41.5 percent below its peak throughput.
FlexServe runs on-device LLM inference inside TrustZone for 4.41 percent more latency
Shanghai Jiao Tong University researchers kept model weights and prompts inside ARM TrustZone at 4.41 percent more time to first token.
NanoSD shrinks Stable Diffusion 1.5 for phone NPUs
NanoSD is a family of 130M to 315M-parameter distilled Stable Diffusion models for image restoration on mobile NPUs.
ExecuTorch 1.0 reaches general availability for on-device PyTorch models
The PyTorch edge runtime promotes Core ML, Qualcomm Hexagon, Arm Ethos-U, Vulkan and XNNPACK backends to production status.
A19 Pro puts Neural Accelerators in every GPU core
Apple says the iPhone 17 Pro chip pairs Neural Accelerators in each of six GPU cores with a 16-core Neural Engine to run large local language models.
ShadowNPU scores attention on the Snapdragon NPU and reports 4.5 times faster inference
Peking University and BUPT move attention token scoring onto a Snapdragon NPU and report up to 4.5 times faster inference using a single CPU core.
P/D-Device prefills in the cloud and decodes on the phone, cutting TTFT 60%
Huawei researchers split prefill and decoding between cloud and phone, reporting time to first token down at least 60% and cloud throughput up to 15x.
Vijay Janapa Reddi argues edge generative AI needs models under 1B parameters
In ACM Queue, Vijay Janapa Reddi argues generative AI reaches edge devices only through small deployable models, with sub-1B ones the viable case.
GenAI at the edge survey lists 12 accelerators, 8 of them only simulated
A Johns Hopkins and Duke survey of generative AI on edge devices puts peak accelerator efficiency at 74.34 TOPS/W, with 8 of 12 designs only simulated.
BlueLM-V-3B runs a multimodal model on a Dimensity 9300 NPU in 2.2 GB
CUHK and vivo AI Lab report 24.4 tok/s and 2.2 GB peak memory for a 3B vision-language model on a MediaTek Dimensity 9300.
PhoneLM searches for a fast architecture before training it and hits 58 tok/s
BUPT researchers picked their 0.5B and 1.5B transformer shapes by measuring speed on a Snapdragon 8 Gen 3 first, then pre-training the winner.