BaseRT

What is BaseRT?

BaseRT is the fastest LLM runtime on Apple Silicon, offering 6.4x performance over llama.cpp and 3.9x over MLX. It enables local LLM inference with zero per-token cost and full privacy, installable via a single command.

Key features

• 6.4x faster than llama.cpp on M-series Macs
• 3.9x faster than Apple MLX framework
• Single-command installation
• No per-token cost — run models locally
• Full data privacy — no data leaves your device
• Supports multiple model formats
• Optimized for always-on background agents
• Open source

Who it's for

• Developers running local LLMs on Apple Silicon for development
• Always-on AI agents that need low-cost continuous inference
• Privacy-sensitive workloads requiring on-device processing
• Building local AI-powered applications without API dependencies
• Testing and benchmarking different open models locally

When not to use it

Not ideal if you need GPU-level performance on non-Apple hardware, or require cloud-scale throughput. BaseRT is optimized specifically for Apple Silicon — cross-platform use cases should consider alternatives like vLLM or TGI.

FAQ