Alternatives to Mirai

Free, MIT-licensed local LLM inference for Apple Silicon Macs, with SSD-backed prefix caching and native OpenAI and Anthropic APIs.

picoLLM runs quantized LLMs fully on-device — local inference for phones, embedded Linux, and microcontrollers with no cloud calls.

Free, open-source local AI server that runs LLMs, image, music, video and 3D generation entirely on your Apple Silicon Mac.

Apple Silicon ML platform for training, serving, quantizing, and merging local models on M1–M5 Macs.

Ollama runs open models locally or through per-token cloud inference, with an OpenAI-compatible API endpoint.

LFM2.5 is Liquid AI's open-weight on-device model family for local text, vision, and speech AI on phones, vehicles, and NPUs.

Downloadable .aimodel bundles and conversion recipes for Apple's Core AI on iOS 27 and macOS 27.

BitNet (bitnet.cpp): Microsoft's open-source 1-bit LLM inference framework for fast, lossless CPU and GPU inference

LLM Hub runs 15+ AI models fully offline on Android, iOS and desktop for private on-device chat, image, video and code generation.

QVAC SDK: one open-source API for on-device AI across text, vision and audio — no cloud, no API keys

Offline AI chat on iPhone, iPad, and Mac that runs open-source LLMs on-device via Core ML — no cloud, no data collection.

Run LLMs, speech-to-text, and image generation fully offline on your device via a Kotlin Multiplatform library, an app, and an IDE plugin.

Stream ONNX weights from disk and run SDXL, Mistral 7B and Whisper on low-RAM devices.

NexaSDK for Mobile runs multimodal AI on-device on iOS and Android via Qualcomm AI Hub.

On-device AI inference SDKs with hand-tuned MetalRT and QHexRT kernels, plus Wally's hosted open-model API.

LM Studio runs open LLMs locally with llama.cpp and MLX, then hands your files to the Bionic agent for document editing, coding and

Self-hosted, MIT-licensed AI runtime that serves text, voice, vision, image, video and 3D workloads through one OpenAI-compatible API.

Atomic Chat runs 1,000+ open models fully offline on Mac, Windows, Linux or phone — no account, no message caps.

Runs open-source AI models entirely on your iPhone or Mac — chat, offline voice, document Q&A, and no account required.

A free desktop app that runs the Mistral 7B model on your own machine, so you can chat with spreadsheets and PDFs without uploading them anywhere.

Open-source JavaScript library that runs small LLMs like Llama 3.2 1B Instruct inside the browser via WebAssembly and WebGPU.

A headless React hooks library that runs Vicuna-13B chat entirely in your browser via WebGPU, with no prompt or completion leaving the client.

Open-source compiler and deployment engine for running LLMs natively on web, mobile, desktop and cloud

Run any GGUF AI model locally on Android with zero network permissions

A free, open-source app that runs small Llama and DeepSeek models entirely on your Apple device, with no internet needed after download.

Iris Android runs GGUF language models offline on your phone using llama.cpp — no server, no account, no data leaving the handset.

RWKV Runner: free open-source desktop app to run and fine-tune infinite-context RWKV models locally

Run local HuggingFace models or plug in your own Claude, GPT and Gemini API keys — free, open-source desktop client

Podman AI Lab runs open source LLMs locally with a model catalog, Kubernetes-ready recipes and OpenAI-compatible endpoints.

DAAAM brings natural-language search to Mac photo and video libraries with on-device AI, so you can find RAW files and clips by