Megrez2 Technical Report

TL;DR

Megrez2 is a lightweight, high-performance language model architecture optimized for device deployment, featuring cross-layer expert sharing and pre-gated routing, achieving competitive results with fewer parameters.

Contribution

Introduction of Megrez2 architecture with cross-layer expert sharing and pre-gated routing, enabling efficient deployment and high performance in resource-constrained environments.

Findings

Megrez2-Preview trained on 5-trillion tokens shows strong performance.

Achieves competitive results with only 3B active parameters.

Demonstrates effectiveness across language understanding, reasoning, and code tasks.

Abstract

We present Megrez2, a novel lightweight and high-performance language model architecture optimized for device native deployment. Megrez2 introduces a novel cross-layer expert sharing mechanism, which significantly reduces total parameter count by reusing expert modules across adjacent transformer layers while maintaining most of the model's capacity. It also incorporates pre-gated routing, enabling memory-efficient expert loading and faster inference. As the first instantiation of the Megrez2 architecture, we introduce the Megrez2-Preview model, which is pre-trained on a 5-trillion-token corpus and further enhanced through supervised fine-tuning and reinforcement learning with verifiable rewards. With only 3B activated and 7.5B stored parameters, Megrez2-Preview demonstrates competitive or superior performance compared to larger models on a wide range of tasks, including language understanding, instruction following, mathematical reasoning, and code generation. These results highlight the effectiveness of the Megrez2 architecture to achieve a balance between accuracy, efficiency, and deployability, making it a strong candidate for real-world, resource-constrained applications.