UV PyPI Mirrors for China — Ranked by Speed & Community Rating
Top Mirrors Ranked
| Rank | Mirror | URL | Throughput | Latency | Community Rating |
|---|---|---|---|---|---|
| 1 | Tsinghua TUNA | https://pypi.tuna.tsinghua.edu.cn/simple |
~3.3 MB/s | ~500ms | 5/5 — Best sync (15min), 99.9% completeness |
| 2 | Alibaba Cloud | https://mirrors.aliyun.com/pypi/simple/ |
~4.8 MB/s | ~300ms | 4/5 — CDN-backed, most stable for production |
| 3 | Huawei Cloud | https://repo.huaweicloud.com/repository/pypi/simple/ |
~13.6 MB/s | Unstable | 3/5 — Fastest throughput, but occasional timeouts |
| 4 | USTC | https://pypi.mirrors.ustc.edu.cn/simple/ |
~0.7 MB/s | ~136ms | 4/5 — Lowest latency, 99.9% completeness |
| 5 | Tencent Cloud | https://mirrors.cloud.tencent.com/pypi/simple/ |
Medium | ~1253ms | 4/5 — Good in South China |
| 6 | Douban | https://pypi.doubanio.com/simple/ |
Medium | Low | 4/5 — Aging, some outdated packages |
| 7 | NetEase | https://mirrors.163.com/pypi/simple/ |
~3.0 MB/s | Medium | 3/5 — Good in North China |
| 8 | Volcano Engine | https://mirrors.volces.com/pypi/simple/ |
Unknown | Unknown | New (ByteDance) |
| 9 | BFSU | https://mirrors.bfsu.edu.cn/pypi/web/simple/ |
Similar to TUNA | Medium | Shares TUNA backend |
| 10 | SJTU | https://mirror.sjtu.edu.cn/pypi/web/simple/ |
Medium | Medium | Good in East China |
Community Consensus
- Tsinghua TUNA — Most recommended overall. 15-min sync frequency, near-complete package coverage. The default recommendation on Zhihu, CSDN, and V2EX.
- Alibaba Cloud — Best for production/commercial use. CDN edge nodes across China give stable speeds nationwide. Fastest in benchmarks that tested from Alibaba's own infrastructure.
- Huawei Cloud — Highest raw throughput in one benchmark (~13.6 MB/s), but unreliable — timed out in latency tests.
- USTC — Lowest latency (~136ms), but lower throughput. Best for metadata-heavy operations.
- Regional picks: South China → Tencent/Alibaba; North China → NetEase/Baidu; East China → Douban/SJTU.
How to Configure UV
Important: UV does NOT read ~/.pip/pip.conf. You must configure it separately.
Option 1: Environment Variables (recommended)
Add to ~/.zshrc:
export UV_DEFAULT_INDEX="https://mirrors.aliyun.com/pypi/simple/"
export UV_INDEX="https://pypi.tuna.tsinghua.edu.cn/simple https://pypi.org/simple"
export UV_PYTHON_INSTALL_MIRROR="https://gh-proxy.com/github.com/indygreg/python-build-standalone/releases/download"
Option 2: Global Config File
Create ~/.config/uv/uv.toml:
[[index]]
url = "https://mirrors.aliyun.com/pypi/simple/"
default = true
Option 3: Project-Level pyproject.toml
[[tool.uv.index]]
url = "https://mirrors.aliyun.com/pypi/simple/"
default = true
Option 4: One-off Command Line
uv pip install numpy --default-index https://mirrors.aliyun.com/pypi/simple/
uv add requests --default-index https://pypi.tuna.tsinghua.edu.cn/simple
Python Installation Mirror
uv python install downloads from GitHub (blocked/slow in China). Set UV_PYTHON_INSTALL_MIRROR to a proxy, or use CERNET's mirror at https://mirrors.cernet.edu.cn/list/python-build-standalone.
Speed Benchmarks
Latency Test (cnpip tool)
| Mirror | Latency |
|---|---|
| USTC | 135.71 ms |
| Alibaba Cloud | 300.77 ms |
| Tsinghua TUNA | 499.51 ms |
| Default PyPI | 1252.75 ms |
| Tencent Cloud | ~1253 ms |
| Huawei Cloud | Timeout |
Download Throughput (1,014 MB package install, China Telecom)
| Mirror | Test 1 | Test 2 | Test 3 | Average |
|---|---|---|---|---|
| Huawei Cloud | 12.0 MB/s | 15.8 MB/s | 13.1 MB/s | ~13.6 MB/s |
| Alibaba Cloud | 5.8 MB/s | 2.6 MB/s | 5.9 MB/s | ~4.8 MB/s |
| Tsinghua TUNA | 8.2 MB/s | 0.5 MB/s | 1.1 MB/s | ~3.3 MB/s |
| NetEase | 2.3 MB/s | 5.5 MB/s | 1.3 MB/s | ~3.0 MB/s |
| USTC | 0.4 MB/s | 1.0 MB/s | 0.7 MB/s | ~0.7 MB/s |
Verdict
Tsinghua + Alibaba as primary/fallback is the safest combo for most users in China. Test from your specific location with cnpip for personalized results, since speed varies heavily by ISP and region.