MobileMoE: Scaling Mixture-of-Experts for Sub-Billion Parameter On-Device Deployment
MobileMoE introduces a family of on-device MoE language models with 0.3–0.9B active parameters and 1.3–5.3B total parameters, targeting mobile deployment under memory and compute constraints. The authors derive an on-device MoE scaling law identifying a sweet spot of moderate sparsity with fine-grained and shared experts, then train models through a four-stage recipe including quantization-aware training on open-source data. Across 14 benchmarks, MobileMoE matches or exceeds leading dense on-device LLMs with 2–4× fewer inference FLOPs, and delivers 1.8–3.8× faster prefill and 2.2–3.4× faster decode than dense baselines on commodity smartphones at comparable INT4 memory.
Related guides (4)
Related events (8)
Complete-muE: Optimal Hyperparameter Transfer and Scaling for MoE Models
Complete-muE is a framework for transferring hyperparameters across dense FFN and Mixture-of-Experts (MoE) transformer architectures, addressing limitations of existing tools like μP and SDE that cannot handle simultaneous architecture and token-per-expert changes. It uses a two-bridge system: Bridge I maps dense FFN to Dense MoE via active-width μP with normalized router scale, and Bridge II maps Dense MoE to sparse MoE via activated-expert scaling with a first-order SDE correction. The practical outcome is a 'tune dense once, transfer to all' recipe that enables near-optimal hyperparameter reuse across MoE configurations without costly re-tuning. Experiments on language model and diffusion model pretraining confirm stable hyperparameter optima across architectures and parameter counts.
ZEDA: Post-Trained MoE Models Can Skip Half Their Experts via Self-Distillation
This paper introduces Zero-Expert Self-Distillation Adaptation (ZEDA), a framework that converts static post-trained Mixture-of-Experts (MoE) language models into dynamic ones without pre-training from scratch. ZEDA injects parameter-free zero-output experts into each MoE layer and uses two-stage self-distillation with the original model as a frozen teacher. Applied to Qwen3-30B-A3B and GLM-4.7-Flash across 11 benchmarks, ZEDA eliminates over 50% of expert FLOPs with marginal accuracy loss and achieves approximately 1.20× end-to-end inference speedup, outperforming the strongest dynamic MoE baseline by 4–6 points.
Expert Tying reduces MoE LLM memory footprint by ~2x with minimal quality loss
Researchers introduce Expert Tying, an architectural modification for Mixture-of-Experts LLMs that shares expert parameters across consecutive transformer layers while keeping routing and attention layer-independent. Evaluated on OLMoE, Qwen3, and DeepSeek-style MoE architectures, the method achieves nearly 2x memory reduction with negligible perplexity or downstream quality degradation. The approach exploits parameter redundancy in MoE pathways to improve the compute-to-memory trade-off for training and inference.
Mixture of Experts Explained
This Hugging Face blog post provides a technical overview of the Mixture of Experts (MoE) architecture, explaining how sparse gating mechanisms route tokens to subsets of expert feed-forward layers to achieve computational efficiency. The post covers training dynamics, inference considerations, and the tradeoffs between dense and sparse models. It serves as a reference document contextualizing MoE's growing relevance following high-profile model releases using the architecture.
EMO: Pretraining Mixture of Experts for Emergent Modularity
AllenAI introduces EMO, a pretraining approach for Mixture of Experts (MoE) models that aims to produce emergent modularity during training. The work explores how MoE architectures can develop specialized expert routing without explicit supervision. Published on the Hugging Face blog, this represents research-level work on improving MoE training dynamics and efficiency.
Global-batch Load Balancing for MoE LLM Training from Qwen
Qwen Research introduces a global-batch load balancing technique for Mixture-of-Experts (MoE) LLM training, claiming it is nearly a 'free lunch' improvement. The method addresses expert load imbalance across training batches, a known efficiency and quality bottleneck in MoE architectures. The approach targets the router and expert activation dynamics in transformer-based MoE layers.
Qwen1.5-MoE: Matching 7B Model Performance with 1/3 Activated Parameters
Alibaba's Qwen team releases Qwen1.5-MoE-A2.7B, a mixture-of-experts model with only 2.7 billion activated parameters that claims performance parity with 7B dense models such as Mistral 7B and Qwen1.5-7B. The model activates roughly one-third of its total parameters during inference, offering significant compute efficiency gains. This release follows growing industry interest in MoE architectures sparked by Mixtral, and the model is available on GitHub, HuggingFace, and ModelScope.
Qwen2.5-Max: Large-Scale MoE Model Release by Alibaba's Qwen Team
Alibaba's Qwen team announces Qwen2.5-Max, a large-scale Mixture-of-Experts language model. The post acknowledges that scaling insights for very large MoE models have been limited, citing DeepSeek V3's recent disclosures as a reference point. The model is positioned as a frontier-scale MoE system developed concurrently with ongoing Qwen2 research.