Large language models (LLMs) power many modern applications, but serving them at scale remains costly and resource-intensive. Current server-centric systems overlook consumer-grade GPUs at the edge. We introduce SpecEdge, an edge-assisted inference framework that splits LLM workloads between edge and server GPUs using a speculative decoding scheme, exchanging only token outputs over the network. SpecEdge employs proactive edge drafting to overlap edge token creation with server verification and pipeline-aware scheduling that interleaves multiple user requests to increase server-side throughput. Experiments show SpecEdge enhances overall cost efficiency by 1.91x through achieving 2.22x server throughput, and reduces inter token latency by 11.24% compared to a server-only baseline, introducing a scalable, cost-effective paradigm for LLM serving.
View on arXiv@article{park2025_2505.17052,
title={ SpecEdge: Scalable Edge-Assisted Serving Framework for Interactive LLMs },
author={ Jinwoo Park and Seunggeun Cho and Dongsu Han },
journal={arXiv preprint arXiv:2505.17052},
year={ 2025 }
}