# Every Expert Counts: ExactMoE for Memory-Efficient W4A16 Inference

arXiv: https://arxiv.org/abs/2608.15383 · HTML: https://mlsanity.com/paper/2608.15383/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Efficient MoE training and inference systems, one of 12 directions in Mixture-of-experts.
- Builds on 21 of its 25 references that are in the map.
- First posted on arXiv 15 Aug 2026.

## Placed in

- Mixture-of-experts / [Efficient MoE training and inference systems](https://mlsanity.com/moe/efficient-moe-training-and-inference-systems/)

## Builds on (most cited first, 21 of 21)

- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [OLMoE: Open Mixture-of-Experts Language Models](https://mlsanity.com/paper/2409.02060/) (2024)
- [MoE-Lightning: High-Throughput MoE Inference on Memory-constrained GPUs](https://mlsanity.com/paper/2411.11217/) (2024)
- [Fiddler: CPU-GPU Orchestration for Fast Inference of Mixture-of-Experts Models](https://mlsanity.com/paper/2402.07033/) (2024)
- [HOBBIT: A Mixed Precision Expert Offloading System for Fast MoE Inference](https://mlsanity.com/paper/2411.01433/) (2024)
- [MARLIN: Mixed-Precision Auto-Regressive Parallel Inference on Large Language Models](https://mlsanity.com/paper/2408.11743/) (2024)
- [MoE-Infinity: Efficient MoE Inference on Personal Machines with Sparsity-Aware Expert Cache](https://mlsanity.com/paper/2401.14361/) (2024)
- [Mixture of Quantized Experts (MoQE): Complementary Effect of Low-bit Quantization and Robustness](https://mlsanity.com/paper/2310.02410/) (2023)
- [MxMoE: Mixed-precision Quantization for MoE with Accuracy and Performance Co-Design](https://mlsanity.com/paper/2505.05799/) (2025)
- [SwapMoE: Serving Off-the-shelf MoE-based Large Language Models with Tunable Memory Budget](https://mlsanity.com/paper/2308.15030/) (2023)
- [MiLo: Efficient Quantized MoE Inference with Mixture of Low-Rank Compensators](https://mlsanity.com/paper/2504.02658/) (2025)
- [Mixture of Experts with Mixture of Precisions for Tuning Quality of Service](https://mlsanity.com/paper/2407.14417/) (2024)
- [Dynamic Expert Quantization for Scalable Mixture-of-Experts Inference](https://mlsanity.com/paper/2511.15015/) (2025)
- [PagedWeight: Efficient MoE LLM Serving with Dynamic Quality-Aware Weight Quantization](https://mlsanity.com/paper/2607.16184/) (2026)

## Abstract (arXiv)

Sparse mixture-of-experts (MoE) language models reduce arithmetic by activating only a small subset of experts per token, yet deployment still requires storing and moving the full expert bank. We present ExactMoE, an inference design that applies symmetric group-128 four-bit weight quantization only to routed experts, stores those experts in kernel-native MARLIN form in pinned host memory, and executes all selected experts through a configurable GPU-resident slot cache and fused grouped MoE kernels. The router, attention, embeddings, normalization layers, and language-model head remain in BF16. "Exact" refers to complete expert availability and an unchanged top-k routing procedure: no expert is pruned, substituted, or forced to execute on the CPU. It does not imply numerical identity with the BF16 model. On OLMoE-1B-7B-0924-Instruct, evaluated on a single NVIDIA L4, a 16-slot configuration reduces peak reserved GPU memory from 14.168 to 1.836 GiB (87.04%) while retaining 81.85% of BF16 decode throughput. A fully resident 64-slot configuration reaches 31.923 tokens/s versus 21.662 tokens/s for BF16 while reserving 4.061 GiB. Across 12,450 zero-shot multiple-choice questions, ExactMoE obtains 70.3534% normalized accuracy versus 70.8996% for BF16, retaining 99.23% of the baseline accuracy. In a matched 16-token ablation, fused grouped execution is 1.97x as fast as a sequential W4 reference. These results identify a practical memory-transfer-throughput frontier for complete-expert MoE inference.