BaseRT is the fastest LLM runtime on Apple Silicon. Install it with one command and run local models on your own device.
Loading comments…
Achievement
Project Info
Product Keywords
BaseRT is a high-performance LLM runtime purpose-built for Apple Silicon. It installs with a single terminal command and lets you run large language models entirely on your own device—no cloud dependency, no API keys, and no data leaving your machine. Designed for speed and local privacy, BaseRT is positioned as the fastest runtime available for Apple’s M-series chips.
BaseRT is installed with a single curl command: curl -LsSf https://basecompute.co/install.sh | sh. No manual compilation, no dependency wrangling—just a shell pipe and you’re ready to serve models.
On decode tasks, BaseRT outperforms MLX by up to 33% and llama.cpp by a similar margin. On prefill tasks, the gap widens dramatically: up to 6.4x faster than llama.cpp and 3.9x faster than MLX, measured on an Apple M5 Pro with models like Qwen3 30B-A3B and Gemma 4 E2B.
BaseRT ships with a plugin that turns your local model into a coding agent. Serve a model with basert serve basecompute/gemma-4-E4B-it, install the plugin, and run pi—everything stays on your machine. No API keys, no data leaving your device.
BaseRT supports a wide range of 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. New models are added through the community and official releases.
BaseRT is the fastest LLM runtime on Apple Silicon—up to 6.4x faster on prefill than llama.cpp.
That performance edge isn’t just a headline number. For developers running coding agents or interactive applications, faster prefill means lower latency on prompt processing, which directly translates to a snappier user experience. The decode speed improvements also mean higher throughput for streaming responses. Combined with the one-command install and local-only operation, BaseRT offers a rare combination of speed and simplicity that’s hard to find in other runtimes.
You work on Apple Silicon and need the fastest possible local inference—especially for coding agents, privacy-sensitive workloads, or any scenario where you want to keep data on-device without sacrificing performance. If you’ve been using MLX or llama.cpp and want to see if a 15–33% decode boost and up to 6.4x faster prefill makes a difference in your workflow, BaseRT is worth a try.
Maker
kettle_dev
Other tools you might consider
Bob's CLI runs on your own hardware with zero API costs, zero data leaving your machine. Bob lives in your terminal, sees your actual files, and writes code only with your explicit approval. What makes it different: auto-detect local AI models, behavioral DNA profiling that adapts to how YOU work, autonomous code review + auto-fix, conversation forking, deep dives, and SovereignLink — remote execution from any device while your code stays home. Free to start. Sovereign by design.
nullsink provides anonymous access to Claude, GPT, and confidential open models without requiring an account, email, card, or subscription. Fund a bearer API key from $2 using Monero or Bitcoin, then use it with native Anthropic or OpenAI-compatible clients. Credit does not expire. nullsink keeps no request-content or client-IP logs and publishes its AGPL source code, architecture, and trust-model limitations.
Run state-of-the-art open-source models (GLM 5.1, Kimi K2.7 Code, MiniMax M2.7, and more) in Claude Code at up to 4× the speed (up to 200 tok/s) for a flat $29/month. Set up in minutes, no code changes.
Today's models are capable enough. Smart enough. Fast enough. But we still feel they don’t fit in the room. Humalike is building the behavioral infrastructure for humanlike AI agents. The social skills & proactiveness your agents have been missing. APIs, models, benchmarks.
Loading comments…
