PublishedSEC 2025

LM-Meter: Unveiling Runtime Inference Latency for On-Device Language Models

Runtime latency inside an on-device language model can be measured live, phase by phase and kernel by kernel, at under 3% throughput cost.

Haoxin Wang (Georgia State University) · Xiaolong Tu (Georgia State University) · Hongyu Ke (Georgia State University) · Huirong Chai (Georgia State University) · Dawei Chen (Toyota InfoTech Labs) · Kyungtae Han (Toyota InfoTech Labs)

VenueThe Tenth ACM/IEEE Symposium on Edge Computing (SEC ’25)
AreasOn-device language model systems
ArtifactsCode

Abstract

Large Language Models (LLMs) are increasingly integrated into everyday applications, but their prevalent cloud-based deployment raises growing concerns around data privacy and long-term sustainability. Running LLMs locally on mobile and edge devices (on-device LLMs) offers the promise of enhanced privacy, reliability, and reduced communication costs. However, realizing this vision remains challenging due to substantial memory and compute demands, as well as limited visibility into performance-efficiency trade-offs on resource-constrained hardware. We propose LM-Meter, the first lightweight, online latency profiler tailored for on-device LLM inference. LM-Meter captures fine-grained, real-time latency at both phase (e.g., embedding, prefill, decode, softmax, sampling) and kernel levels without auxiliary devices. We implement LM-Meter on commercial mobile platforms and demonstrate its high profiling accuracy with minimal system overhead, e.g., only 2.58% throughput reduction in prefill and 0.99% in decode under the most constrained Powersave governor. Leveraging LM-Meter, we conduct comprehensive empirical studies revealing phase- and kernel-level bottlenecks in on-device LLM inference, quantifying accuracy-efficiency trade-offs, and identifying systematic optimization opportunities.

2.58%

Throughput reduction in prefill, under the most constrained governor

0.99%

Throughput reduction in decode

SEC ’25

ACM/IEEE Symposium on Edge Computing

The problem

Local inference is a black box

5

Inference phases profiled: embedding, prefill, decode, softmax, sampling

Kernel

Granularity, without auxiliary measurement hardware

Cite this work

Haoxin Wang, Xiaolong Tu, Hongyu Ke, Huirong Chai, Dawei Chen, and Kyungtae Han (2025). LM-Meter: Unveiling Runtime Inference Latency for On-Device Language Models. The Tenth ACM/IEEE Symposium on Edge Computing (SEC ’25). https://doi.org/10.1145/3769102.3770614

@inproceedings{wang2025lmmeter,
  title = {LM-Meter: Unveiling Runtime Inference Latency for On-Device Language Models},
  author = {Haoxin Wang and Xiaolong Tu and Hongyu Ke and Huirong Chai and Dawei Chen and Kyungtae Han},
  booktitle = {The Tenth ACM/IEEE Symposium on Edge Computing (SEC ’25)},
  year = {2025},
  doi = {10.1145/3769102.3770614},
  url = {https://doi.org/10.1145/3769102.3770614}
}