# MoDM: Efficient Serving for Image Generation via Mixture-of-Diffusion Models

arXiv: https://arxiv.org/abs/2503.11972 · HTML: https://mlsanity.com/paper/2503.11972/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 15 Mar 2025.

## Built on by (most cited first, 12 of 12)

- [TetriServe: Efficiently Serving Mixed DiT Workloads](https://mlsanity.com/paper/2510.01565/) (2025)
- [HybridStitch: Pixel and Timestep Level Model Stitching for Diffusion Acceleration](https://mlsanity.com/paper/2603.07815/) (2026)
- [Speculative Decoding for Autoregressive Video Generation](https://mlsanity.com/paper/2604.17397/) (2026)
- [DiLaServe: High SLO Attainment Serving for Diffusion Language Models](https://mlsanity.com/paper/2606.29094/) (2026)
- [DSTAR: Accelerating Diffusion Transformers via Spatial and Temporal Redundancy Reduction](https://mlsanity.com/paper/2607.15846/) (2026)
- [Attacks on Approximate Caches in Text-to-Image Diffusion Models](https://mlsanity.com/paper/2508.20424/) (2025)
- [Sphinx: Efficiently Serving Novel View Synthesis using Regression-Guided Selective Refinement](https://mlsanity.com/paper/2511.18672/) (2025)
- [Context-Aware Initialization for Reducing Generative Path Length in Diffusion Language Models](https://mlsanity.com/paper/2512.19004/) (2025)
- [StreamWise: Serving Multi-Modal Generation in Real-Time at Scale](https://mlsanity.com/paper/2603.05800/) (2026)
- [SoundWeaver: Semantic Warm-Starting for Text-to-Audio Diffusion Serving](https://mlsanity.com/paper/2603.07865/) (2026)
- [FlashDiff: Efficient Regional Execution and Scheduling for Diffusion Model Serving](https://mlsanity.com/paper/2607.12121/) (2026)
- [rMuscle: Robotic Muscle Memory for Efficient Vision-Language-Action Model Inference](https://mlsanity.com/paper/2609.19104/) (2026)

## Abstract (arXiv)

Diffusion-based text-to-image generation models trade latency for quality: small models are fast but generate lower-quality images, while large models produce better images but are slow. We present MoDM, a novel caching-based serving system for diffusion models that dynamically balances latency and quality through a mixture of diffusion models. Unlike prior approaches that rely on model-specific internal features, MoDM caches final images, allowing seamless retrieval and reuse across multiple diffusion model families. This design enables adaptive serving by dynamically balancing latency and image quality: using smaller models for cache-hit requests to reduce latency while reserving larger models for cache-miss requests to maintain quality. Small model image quality is preserved using retrieved cached images. We design a global monitor that optimally allocates GPU resources and balances inference workload, ensuring high throughput while meeting service-level objectives under varying request rates. Our evaluations show that MoDM significantly reduces average serving time by 2.5x while retaining image quality, making it a practical solution for scalable and resource-efficient model deployment.