Mooncake is a high-performance, distributed key-value (KV) caching system optimized for large-scale large language model (LLM) inference. The Mooncake image series integrates core inference components — the Mooncake Transfer Engine, SGLang, and vLLM — to support efficient model service deployment.
Image list
| Release date | Registry address |
|---|---|
| January 2026 | ac2-registry.cn-hangzhou.cr.aliyuncs.com/ac2/mooncake:0.3.8-sglang0.5.7-vllm0.13.0-pytorch2.9.1-cuda12.9.1-py312-ubuntu24.04 |
| December 2025 | ac2-registry.cn-hangzhou.cr.aliyuncs.com/ac2/mooncake:0.3.7.post2-sglang0.5.5.post3-vllm0.11.0-pytorch2.8.0-cuda12.9.1-py312-ubuntu24.04 |
Release notes
Mooncake 0.3.8 (January 2026)
-
Release date: January 2026
-
Registry address:
ac2-registry.cn-hangzhou.cr.aliyuncs.com/ac2/mooncake:0.3.8-sglang0.5.7-vllm0.13.0-pytorch2.9.1-cuda12.9.1-py312-ubuntu24.04
Core updates
All core components are updated to their latest upstream versions. A new etcd integration simplifies cluster coordination for Mooncake deployments.
Component updates
| Component | Previous version | Current version | What changed |
|---|---|---|---|
| Mooncake Transfer Engine | 0.3.7.post2 | 0.3.8 | Latest upstream release |
| SGLang | 0.5.5.post3 | 0.5.7 | Latest upstream release |
| vLLM | 0.11.0 | 0.13.0 | Latest upstream release |
| PyTorch | 2.8.0 | 2.9.1 | Latest upstream release |
| elastic Remote Direct Memory Access (eRDMA) | 51.2 | 56.2 | Latest upstream release |
New components
| Component | Version | Description |
|---|---|---|
| etcd | 3.6.7 | Distributed key-value store that simplifies cluster coordination for Mooncake deployments |
Component versions
| Component | Version |
|---|---|
| Mooncake Transfer Engine | 0.3.8 |
| SGLang | 0.5.7 |
| sgl-kernel | 0.3.20 |
| sgl-model-gateway | 0.3.0 |
| vLLM | 0.13.0 |
| PyTorch | 2.9.1 |
| CUDA | 12.9.1 |
| NVSHMEM | 3.4.5-1 |
| ERDMA | 56.2-1.0.1 |
| DeepEP | 9af0e0d |
| etcd | 3.6.7 |
| Python | 3.12 |
| Base OS | Ubuntu 24.04 |
Known issues
No known issues in this release.
Mooncake 0.3.7.post2 (December 2025)
-
Release date: December 2025
-
Registry address:
ac2-registry.cn-hangzhou.cr.aliyuncs.com/ac2/mooncake:0.3.7.post2-sglang0.5.5.post3-vllm0.11.0-pytorch2.8.0-cuda12.9.1-py312-ubuntu24.04
Core updates
This is the first monthly Mooncake image. It supports SGLang and vLLM inference engines and enables the Alibaba Cloud eRDMA feature. All core components are updated to their latest upstream versions.
Component versions
| Component | Version |
|---|---|
| Mooncake Transfer Engine | 0.3.7.post2 |
| SGLang | 0.5.5.post3 |
| sgl-kernel | 0.3.17.post1 |
| vLLM | 0.11.0 |
| PyTorch | 2.8.0 |
| CUDA | 12.9.1 |
| NVSHMEM | 3.4.5-1 |
| ERDMA | 51.2-1.0.1 |
| DeepEP | 9af0e0d |
| Python | 3.12 |
| Base OS | Ubuntu 24.04 |
Known issues
No known issues in this release.
Image disclaimer
-
Mooncake images are updated monthly. Components use the latest compatible versions from the upstream community.
-
All components in the Mooncake images are sourced from upstream communities. Alibaba Cloud does not guarantee the functionality or performance of these components.