Show HN: We made LM studio alternative based on own engine (trymirai.com)

A developer posted a Show HN project: a lightweight LM Studio / Ollama alternative built natively for macOS and Apple Silicon that runs models locally with an in-house inference engine. The app bills itself as faster and simpler than existing desktop runtimes, emphasizing complete privacy and security by keeping inference on-device. It lists support for a range of models and sources — e.g., Gemma, Polaris, Llama, Qwen, Hugging Face and DeepSeek — and aims to provide a chat-style interface for interacting with those models directly on your Mac. For the AI/ML community this matters because a native, Apple‑optimized runtime can reduce latency and energy use compared with cross-platform or containerized solutions, while lowering the friction for privacy‑sensitive local workflows. Using a bespoke engine suggests the team has implemented specific performance optimizations and model format interoperability, which could broaden desktop access to many open and commercial models. Important follow-ups to watch: exact hardware acceleration (Metal/ANE) and quantization strategies, model format and license support, and how updates and security patches will be handled — all of which determine real-world performance, compatibility, and trust for developers and users.