[ICLR 2026] MeanCache: From Instantaneous to Average Velocity for Accelerating Flow Matching Inference
Yuren You2, Ting Lu1,2, Chao Tan1,2, Shaoan Zhao1,2, Zhaoxiang Liu1,2
Fang Zhao1,2*, Kai Wang1,2, Shiguo Lian1,2*
2Unicom Data Intelligence, China Unicom,
3National Key Laboratory for Novel Software Technology, Nanjing University
MeanCache_720p.mp4
In Flow Matching inference, existing caching methods primarily rely on reusing Instantaneous Velocity or its feature-level proxies. However, we observe that instantaneous velocity often exhibits sharp fluctuations across timesteps. This leads to severe trajectory deviations and cumulative errors, especially as the cache interval increases. Inspired by MeanFlow, we propose MeanCache. Compared to unstable instantaneous velocity, Average Velocity is significantly smoother and more robust over time. By shifting the caching perspective from a single "point" to an "interval," MeanCache effectively mitigates trajectory drift under high acceleration ratios.
- [2026/02/05] Community Contribution: ComfyUI-MeanCache-Z is now available! thanks to @facok!
- [2025/02/04] Support Z-Image and released the MeanCache vs. LeMiCa comparative study.
- [2025/02/02] Support Qwen-Image and Inference Code Released !
This benchmark evaluates the performance of MeanCache against LeMiCa using the Qwen-Image-2512 model as the base.
Baseline Latency (Original Qwen-Image-2512): 32.8s
| Constraint | Method | Latency | Speedup | Time Reduction |
|---|---|---|---|---|
| LeMiCa | 18.83 s | 1.74x | - | |
| MeanCache | 17.13 s | 1.91x | 9.0% | |
| LeMiCa | 14.35 s | 2.29x | - | |
| MeanCache | 11.67 s | 2.81x | 18.7% | |
| LeMiCa | 10.41 s | 3.15x | - | |
| MeanCache | 6.95 s | 4.72x | 33.2% |
| Constraint | Method | PSNR (↑) | SSIM (↑) | LPIPS (↓) |
|---|---|---|---|---|
| LeMiCa | 29.20 | 0.945 | 0.065 | |
| MeanCache | 29.46 | 0.944 | 0.057 | |
| LeMiCa | 24.31 | 0.835 | 0.176 | |
| MeanCache | 26.49 | 0.907 | 0.104 | |
| LeMiCa | 17.80 | 0.637 | 0.368 | |
| MeanCache | 19.44 | 0.767 | 0.237 |
| Z-Image-base | MeanCache(B=25) | MeanCache(B=20) | MeanCache(B=15) | MeanCache(B=13) |
|---|---|---|---|---|
| 18.07 s | 9.15 s | 7.36 s | 5.58 s | 4.85 s |
 |
 |
 |
 |
 |
| Method | Qwen-Image-2512 | MeanCache(B=25) | MeanCache(B=17) | MeanCache(B=10) |
|---|---|---|---|---|
| Latency | 32.8 s | 17.13 s | 11.67 s | 6.95 s |
| T2I |  |
 |
 |
 |
| Method | Qwen-Image | MeanCache(B=25) | MeanCache(B=17) | MeanCache(B=10) |
|---|---|---|---|---|
| Latency | 33.13 s | 17.04 s | 11.63 s | 6.92 s |
| T2I |  |
 |
 |
 |
The majority of this project is released under the Apache 2.0 license as found in the LICENSE file.
If you find MeanCache useful in your research or applications, please consider giving us a star ⭐ and citing it by the following BibTeX entry:
@inproceedings{gao2025meancache,
title = {MeanCache: From Instantaneous to Average Velocity for Accelerating Flow Matching Inference},
author = {Huanlin Gao and Ping Chen and Fuyuan Shi and Ruijia Wu and Yantao Li and Qiang Hui and Yuren You and Ting Lu and Chao Tan and Shaoan Zhao and Zhaoxiang Liu and Fang Zhao and Kai Wang and Shiguo Lian},
journal = {International Conference on Learning Representations (ICLR)},
year = {2026},
url = {https://arxiv.org/abs/2601.19961}
}