Apple Silicon with Metal

​

Prerequisites

sudo xcode-select -s /Applications/Xcode.app/Contents/Developer

​

Install SGLang

​

Install from Source

# Use the default branch
git clone https://github.com/sgl-project/sglang.git
cd sglang

# Create and activate a virtual environment
uv venv -p 3.12 sglang-metal
source sglang-metal/bin/activate

# (Optional) Compile sgl-kernel
uv pip install --upgrade pip
uv run python/sglang/kernels/aot/setup_metal.py install

# Install sglang python package along with diffusion support
rm -f python/pyproject.toml && mv python/pyproject_other.toml python/pyproject.toml
uv pip install -e "python[all_mps]"

​

Launch of the Serving Engine

SGLANG_USE_MLX=1 python -m sglang.launch_server \
  --model <MODEL_ID_OR_PATH> \
  --disable-cuda-graph \
  --host 0.0.0.0

1. SGLANG_USE_MLX=1 - Enables the use of MLX as the SGLang runtime backend (if disabled, SGLang will fall back to torch.mps, which has less support)
2. --disable-cuda-graph - Disables usage of CUDA graph, which is not relevant for Apple Metal.
3. --disable-overlap-schedule - Disables overlap scheduling (enabled/not present by default) achieved using MLX’s async_eval()
4. SGLANG_MLX_USE_CUSTOM_ROPE=1 - Enables the optional custom Metal RoPE kernel. It is disabled by default, so the MLX backend uses the standard RoPE path unless you opt in for A/B testing.
5. SGLANG_MLX_FUSE_SWIGLU=1 - Enables the use of fused Swish-Gated Linear Unit kernel (disabled by default)
6. SGLANG_MLX_CLEAR_CACHE_STEPS=256 - Sets the number of decode steps before clearing the MLX cache (256 by default)

​

Quantization

1. Pre-quantized HF repos. Any mlx-community/<model>-4bit (or -8bit) repo loads directly through mlx_lm.load(...) — no extra flag needed.
   SGLANG_USE_MLX=1 python -m sglang.launch_server \
     --model-path mlx-community/Qwen3-0.6B-4bit \
     --disable-cuda-graph
   
2. On-the-fly quantization. For any fp16 model, pass --quantization mlx_q4 or --quantization mlx_q8 to have sglang quantize the weights at load time via mlx_lm.utils.quantize_model (group size 64, the mlx-community default). The quantized weights stay in process memory; the on-disk model is untouched.
   SGLANG_USE_MLX=1 python -m sglang.launch_server \
     --model-path Qwen/Qwen3-0.6B \
     --quantization mlx_q4 \
     --disable-cuda-graph
   
   Expected log line:
   Quantizing MLX model on-the-fly: bits=4 group_size=64 (preset=mlx_q4)
   Quantization complete in 0.13s — active mem: 1.11 GB -> 0.31 GB (71.9% reduction)
   
   The MLX backend silently ignores --quantization mlx_q4 when the model is already quantized in its HF config (path 1), so the same flag is safe to pass either way.

​

Benchmarking with Requests

​

Throughput Testing

SGLANG_USE_MLX=1 python -m sglang.benchmark.one_batch \
  --model-path <MODEL_ID_OR_PATH> \
  --disable-cuda-graph \
  --tp-size 1 \
  --batch-size 1 \
  --input-len 60 \
  --output-len 10

SGLANG_USE_MLX=1 python -m sglang.benchmark.offline_throughput \
  --model-path <MODEL_ID_OR_PATH> \
  --disable-cuda-graph \
  --num-prompts 1 \
  --disable-overlap-schedule

SGLANG_USE_MLX=1 python -m sglang.benchmark.offline_throughput \
  --model-path <MODEL_ID_OR_PATH> \
  --disable-cuda-graph \
  --num-prompts 1