Skip to main content

SGLang Installation (PyPI)

This guide covers installing SGLang via PyPI using pip or uv. For Docker-based installation, refer to the individual model deployment guides.

For the full installation reference, see the official SGLang installation guide.

Stable Releases​

With CUDA-specific kernels​

CUDA 12.9​

Using uv:

uv pip install sglang sgl-kernel \
  --extra-index-url https://sgl-project.github.io/whl/cu129/ \
  --extra-index-url https://download.pytorch.org/whl/cu129 \
  --index-strategy unsafe-best-match

Using pip:

pip install sglang sgl-kernel \
  --extra-index-url https://sgl-project.github.io/whl/cu129/ \
  --extra-index-url https://download.pytorch.org/whl/cu129

CUDA 13.0​

Using uv:

uv pip install sglang sgl-kernel \
  --extra-index-url https://sgl-project.github.io/whl/cu130/ \
  --extra-index-url https://download.pytorch.org/whl/cu130 \
  --index-strategy unsafe-best-match

Using pip:

pip install sglang sgl-kernel \
  --extra-index-url https://sgl-project.github.io/whl/cu130/ \
  --extra-index-url https://download.pytorch.org/whl/cu130

Nightly Releases​

Nightly builds include the latest features and model support before they land in a stable release. Some recently added models (e.g., Ling-2.5-1T) are available via nightly builds before stable release.

CUDA 12.9​

Using uv:

uv pip install -U sglang sgl-kernel --pre \
  --index-url https://sgl-project.github.io/whl/cu129/ \
  --extra-index-url https://pypi.org/simple \
  --extra-index-url https://download.pytorch.org/whl/cu129 \
  --index-strategy unsafe-best-match

Using pip:

pip install -U sglang sgl-kernel --pre \
  --index-url https://sgl-project.github.io/whl/cu129/ \
  --extra-index-url https://pypi.org/simple \
  --extra-index-url https://download.pytorch.org/whl/cu129

CUDA 13.0​

Using uv:

# Step 1: Install nightly sglang
uv pip install -U sglang --pre \
  --index-url https://sgl-project.github.io/whl/cu129/ \
  --extra-index-url https://pypi.org/simple \
  --extra-index-url https://download.pytorch.org/whl/cu130 \
  --index-strategy unsafe-best-match

# Step 2: Install CUDA 13.0 kernel
uv pip install -U sgl-kernel \
  --extra-index-url https://sgl-project.github.io/whl/cu130/ \
  --extra-index-url https://download.pytorch.org/whl/cu130 \
  --index-strategy unsafe-best-match