macOS support

Langflow runs on Apple Silicon and Intel Macs. Local features that depend on PyTorch are available on Apple Silicon only. PyTorch publishes no wheels for Intel Macs, so Langflow skips those extras there.

Quick reference

Feature	Apple Silicon	Intel Mac
Core Langflow	Supported	Supported
PyTorch-based components	Supported	Not available
Metal GPU acceleration	Supported	Not available
Native OCR	Supported	Supported

Python versions

Langflow supports Python 3.10 through 3.14 on both Apple Silicon and Intel Macs.

On every Intel Mac, including Python 3.10 through 3.12, Langflow skips extras that depend on PyTorch. PyTorch 2.2.2 was the last release with Intel Mac wheels, and Langflow does not install that release.

What works on both

These features work on Apple Silicon and Intel Macs:

• Flow builder and visual editor
• API server and endpoints
• Database operations with SQLite and PostgreSQL
• Authentication and user management
• Components that do not run a local model, including text processing, API calls, and data transformations
• Native OCR through ocrmac, which uses the macOS Vision framework
• Vector stores and embeddings when the provider is an API
• LangChain components that do not run a local model

Apple Silicon only

Langflow skips these extras on Intel Macs:

• HuggingFace Transformers, for local models. The HuggingFace Inference API still works from an Intel Mac.
• EasyOCR
• Docling local document conversion. docling-core, which handles metadata only, still installs on Intel Macs.

Two more extras install only on some Apple Silicon setups:

• MLX and MLX-VLM install on Apple Silicon with Python 3.12 or newer.
• CUGA installs on Apple Silicon with Python below 3.14. CUGA is also available on Linux and Windows with Python below 3.14.

The sentence-transformers extra has no Intel skip marker. Installing it on an Intel Mac fails when the installer tries to download PyTorch.

Workarounds for Intel Mac users

To use models from an Intel Mac:

1. Call an API provider, such as OpenAI, Cohere, Anthropic, or the HuggingFace Inference API.

2. Run Langflow on a machine where PyTorch installs, and call that instance from the Intel Mac.

3. For some workloads, run the Apple Silicon container image under Rosetta 2:

   docker run --platform linux/arm64 langflowai/langflow:latest

   This is slower than a native Apple Silicon install, and not every feature works.

Installation

On Apple Silicon, install Langflow with the usual command:

uv pip install langflow

To add local ML extras:

uv pip install "langflow[langchain-huggingface,easyocr]"

On an Intel Mac, install the base package:

uv pip install langflow

The HuggingFace, EasyOCR, and Docling local-conversion extras are skipped. The sentence-transformers extra is not skipped, and that install fails because PyTorch is unavailable.

See also

For install errors, see Troubleshoot Intel Mac issues.

For more help, ask in GitHub Discussions or on the Discord server.