BaseRT

BaseRT

BaseRT by BaseCompute is the fastest LLM runtime for Apple Silicon, enabling one-command installation and local model execution on your device.

Apple SiliconLLM runtimelocal inferenceBaseComputemacOS AI
Open site
image for BaseRTOpen site

What is BaseRT?

BaseRT is a high-performance runtime developed by BaseCompute, designed specifically for Apple Silicon hardware. It enables users to install and run large language models (LLMs) locally on their device with a single command. The tool is marketed as the fastest LLM runtime on Apple Silicon, offering significant speed improvements over alternatives like MLX and llama.cpp. Users can serve models directly from their machine, keeping all data on-device without requiring API keys or external network calls.

Application scenarios

  • Local AI development

    Run and test open-source models entirely on your Apple Silicon Mac without cloud dependencies.

  • Coding agents

    Serve a model locally and point a coding agent at it, enabling private, offline code assistance.

  • Privacy-sensitive tasks

    Execute AI workloads without sending data off your device, ideal for confidential or sensitive projects.

  • Model benchmarking

    Compare performance across models using BaseRT’s built-in benchmarks against MLX and llama.cpp.

  • Community experimentation

    Join a Discord community for engineers working on on-device AI and open-source models.

Core Features

  • One-command installation

    Install BaseRT with a single `curl` command: `curl -LsSf https://basecompute.co/install.sh | sh`.

  • Fastest Apple Silicon runtime

    Achieves up to 33% faster decode speeds and up to 6.4x faster prefill speeds compared to llama.cpp, and up to 3.9x faster prefill than MLX.

  • Model serving

    Serve a model locally using `basert serve basecompute/gemma-4-E4B-it` for immediate access.

  • Coding agent integration

    Install a dedicated plugin (`pip install git:github.com/basecompute/pi-basert`) to run local coding agents with BaseRT.

  • Supported models

    Works with popular open-source models including Qwen3, Qwen3.5, Qwen3.6, Llama 3.1, Llama 3.2, Gemma 3, Gemma 4, Mistral, Phi-3, and Nomic BERT.

  • Community support

    Access technical reports, documentation, and a GitHub repository, plus a Discord server for engineers building on-device AI.

Target users

BaseRT is built for engineers and developers working with on-device AI, particularly those who need to run open-source models locally on Apple Silicon hardware. It’s ideal for AI researchers, coding tool developers, and privacy-conscious users who want to avoid cloud dependencies. The tool also appeals to hobbyists and community members interested in benchmarking and experimenting with LLMs on Macs.

How to use BaseRT?

  1. Open a terminal on your Apple Silicon Mac.
  2. Run the installation command: curl -LsSf https://basecompute.co/install.sh | sh.
  3. Serve a model locally using basert serve basecompute/gemma-4-E4B-it (or any supported model).
  4. For coding agents, install the plugin with pip install git:github.com/basecompute/pi-basert and run it.
  5. Access detailed technical reports and documentation via the BaseRT website or GitHub.

Effect review

BaseRT delivers on its promise of being the fastest LLM runtime for Apple Silicon, with concrete benchmark numbers showing substantial gains over MLX and llama.cpp in both decode and prefill phases. The one-command installation and straightforward model serving make it accessible for developers who want to quickly spin up local AI without complex setup. The integration with coding agents is a practical feature for those building private, offline development tools. While the tool is clearly optimized for Apple Silicon, its focus on local execution and privacy aligns well with current trends in edge AI. The active community on Discord and available technical reports suggest a well-supported project, though real-world user feedback beyond benchmarks is not provided on the site.

Frequently Asked Questions

What is BaseRT?
BaseRT is a high-performance LLM runtime optimized for Apple Silicon, allowing you to install and run large language models locally on your device with a single command.
Which Apple devices are supported?
BaseRT supports all Apple Silicon Macs, including M1, M2, M3, and later chips. It is not compatible with Intel-based Macs.
How do I install BaseRT?
Installation is one-command: open Terminal and run the provided install script from BaseCompute's official repository. No complex setup is required.
Is BaseRT free to use?
Yes, BaseRT is free and open-source. You can download and use it without any subscription or licensing fees.
What models can I run with BaseRT?
BaseRT supports a wide range of open-source LLMs, including LLaMA, Mistral, and Phi, optimized for local inference on Apple Silicon.
Does BaseRT require an internet connection?
No, after initial installation and model download, BaseRT runs entirely offline, ensuring privacy and low latency.

BaseRT - AI Tool Detail

BaseRT by BaseCompute is the fastest LLM runtime for Apple Silicon, enabling one-command installation and local model execution on your device.

Category：Training Deployment Tool

Visit Link：https://www.basecompute.co/getbasert

Tags：Apple Silicon、LLM runtime、local inference、BaseCompute、macOS AI