# Full-Pipeline Inference Optimization for MiMo-V2.5 Series: Pushing Hybrid SWA Efficiency to the Limit

arXiv: https://arxiv.org/abs/2607.13095 · HTML: https://mlsanity.com/paper/2607.13095/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Efficient MoE training and inference systems, one of 12 directions in Mixture-of-experts.
- Builds on 1 of its 4 references that are in the map.
- First posted on arXiv 14 Jul 2026.

## Placed in

- Mixture-of-experts / [Efficient MoE training and inference systems](https://mlsanity.com/moe/efficient-moe-training-and-inference-systems/)
- Efficient LLM inference and serving / [KV cache optimization](https://mlsanity.com/inference/kv-cache-optimization/)

## Builds on (most cited first, 1 of 1)

- [SGLang: Efficient Execution of Structured Language Model Programs](https://mlsanity.com/paper/2312.07104/) (2023)

## Built on by (most cited first, 3 of 3)

- [Deeply Interleaved Text-Image Contexts for Multimodal LLMs Assessment](https://mlsanity.com/paper/2609.02573/) (2026)
- [PDEU-Bench: Benchmarking the Personalized Planning Lifecycle of Tool-Calling LLM Agents](https://mlsanity.com/paper/2609.34930/) (2026)
- [ExperienceIndex: Artifact-Grounded Memory](https://mlsanity.com/paper/2610.10091/) (2026)

## Abstract (arXiv)

We present a full-pipeline inference optimization for the MiMo-V2.5 model family, which combines Hybrid Sliding Window Attention (Hybrid SWA), sparse Mixture-of-Experts (MoE), and multimodal encoders. While Hybrid SWA can ideally reduce both attention compute and KVCache storage significantly compared to Full Attention, realizing these gains in production requires substantial engineering effort. We systematically optimize the KVCache system with layerwise prefetch, SWA-aware prefix cache trees, and specialized placement strategies, achieving strict O(W) SWA storage and high cache hit rates. We further build GCache, a high-performance distributed cache infrastructure with RDMA-optimized networking, and develop a KVCache-affinity router to reduce computation while preserving load balancing. We also optimize for multimodal inputs, including GPU image preprocessing, parallel video decoding, and multimodal cache sharing. Together, these optimizations constitute the first large-scale LLM serving system in production that efficiently covers the Hybrid SWA + MoE + multimodal composite architecture.