FreeToken: Efficient Edge-Native MoE Serving with Bandwidth-Adaptive Execution

2026-08-17

AI Summary

This study addresses the challenges of deploying frontier Mixture-of-Experts (MoE) models on edge devices amid resource heterogeneity by proposing an edge-native MoE serving system. Through full-stack co-design, we introduce a dynamic resource mapping mechanism that overcomes the limitations of static offloading strategies. By integrating model layout optimization, expert residency, and CPU-GPU collaborative execution, the system achieves bandwidth adaptivity and efficient runtime memory management. Supporting over 20 architectures, it enables elastic inference for models ranging from 35B to 753B parameters across heterogeneous devices, from laptops to workstations. Consequently, this approach effectively transforms personal computing devices into flexible inference platforms capable of handling agent-centric workloads, significantly enhancing the accessibility of large-scale MoE models in resource-constrained edge environments.

Abstract

Frontier open-weight models are increasingly available, but serving them still largely assumes datacenter infrastructure. We present FreeToken, an edge-native MoE serving system that treats a personal machine not as a small GPU, but as a unified, elastic inference platform. FreeToken co-designs the full serving stack, including model layout and loading, expert residency, CPU--GPU execution, agentic state reuse, and runtime memory management, around two realities of local AI: agent workloads continuously change their execution pattern, and edge hardware exposes heterogeneous resources whose balance differs from machine to machine. Rather than committing to a fixed offloading strategy, FreeToken continuously maps computation and model state onto the resources actually available. FreeToken supports more than 20 MoE models and real coding and tool-using agents across hardware ranging from an 8GB laptop GPU to a single workstation GPU. More importantly, it changes what these machines can practically serve, from a 35B model on a laptop to a 284B model on a gaming desktop and the 753B GLM-5.2 on a single workstation GPU. FreeToken turns open weights into deployable local software, making the machines users already own a practical platform for frontier-scale intelligence. We release the system at flashml.ai.

Problem

• Edge-native MoE Serving
• Heterogeneous Resources
• Agent Workloads
• Open-weight Models
• Local AI Deployment

Innovation

• Edge-Native MoE Serving
• Bandwidth-Adaptive Execution
• Heterogeneous Resource Management
• Full-Stack Co-design
• Agentic State Reuse