LASER optimizes edge deployment of reasoning LLMs with 38% lower latency

Large reasoning models like DeepSeek-R1 achieve strong performance through extended chain-of-thought (CoT) generation, but deploying them on edge devices creates a conflict between long CoT sequences and constrained resources. Existing confidence-based early exit methods use fixed thresholds designed for single requests, ignoring multi-request concurrency and load fluctuations in real-world edge serving. To bridge this gap, researchers from multiple institutions propose LASER (Load-Aware Serving with Early-Exit for Reasoning LLMs at the Edge). LASER couples two complementary designs: a load-aware adaptive exit threshold that adjusts the confidence bar based on real-time system load within an empirically validated robust range, and a difficulty- and load-aware reasoning budget pre-allocation that assigns compute resources by request difficulty and system capacity. The problem is formulated as a joint optimization of reasoning quality and service latency.

Experiments on two reasoning models, four benchmarks, and diverse load conditions show that LASER reduces average latency by 17–38% and improves service-level objective (SLO) satisfaction by 3–6% over fixed-threshold baselines, at an average accuracy cost of only 1%. This makes LASER a practical solution for deploying advanced reasoning AI on resource-constrained edge devices—such as smartphones, IoT hubs, and local servers—where real-time responses are critical. The work is set to appear at WASA 2026 and offers a scalable path for bringing high-performing LLMs to the edge without sacrificing responsiveness.