Back to Blog
May 15, 2026 • Sarah Perez

Osaurus: Local and Cloud AI Models for Mac

Osaurus: Local and Cloud AI Models for Mac Osaurus: Local and Cloud AI Models for Mac Osaurus: Local and Cloud AI Models for Mac

Osaurus brings both local and cloud AI models to your Mac

Osaurus is an open-source, Apple-only LLM server that allows users to seamlessly switch between local and cloud AI models while keeping their files, tools, and model memory on their own hardware.

Overview

What is Osaurus? Osaurus evolved from Dinoki, an AI-powered desktop companion concept ("AI-powered Clippy"). After users questioned why they should pay for an app that still required token costs, co-founder Terence Pae (ex-Tesla, ex-Netflix) pivoted to creating a local AI solution.

Core Functionality:

  • Flexibly connects with locally hosted AI models or cloud providers (OpenAI, Anthropic)
  • Users can freely choose which AI models they're using
  • Keeps models' memory, files, and tools on user's own hardware
  • Functions as a "harness" — a control layer connecting different AI models, tools, and workflows through a single interface

Key Features

Model Support:

  • Local Models: MiniMax M2.5, Gemma 4, Qwen3.6, GPT-OSS, Llama, DeepSeek V4, Apple's on-device foundation models, Liquid AI's LFM family
  • Cloud Providers: OpenAI, Anthropic, Gemini, xAI/Grok, Venice AI, OpenRouter, Ollama, LM Studio

Technical Capabilities:

  • Full MCP (Model Context Protocol) server
  • Ships with 20+ native plug-ins: Mail, Calendar, Vision, macOS Use, XLSX, PPTX, Browser, Music, Git, Filesystem, Search, Fetch, and more
  • Voice capabilities
  • Hardware-isolated, virtual sandbox for security

Technical Requirements

  • Minimum: 64GB RAM
  • Recommended for larger models (DeepSeek v4): 128GB RAM
  • Platform: Mac only

Competitive Advantages

Differentiation from competitors (Ollama, Msty, LM Studio):

  • More user-friendly interface for non-developers
  • Unlike OpenClaw or Hermes (which target developers with terminal interfaces), Osaurus presents an easy-to-use consumer interface
  • Addresses security concerns with hardware-isolated virtual sandbox
  • Different AI models have different strengths; Osaurus lets users switch to the model that best fits their needs

Current Status & Future Direction

Traction:

  • Downloaded over 112,000 times since launch (nearly a year ago)
  • Open source project built in public

Team:

  • Co-founders: Terence Pae (Tesla, Netflix background) and Sam Yoo
  • Currently participating in Alliance accelerator (New York-based)

Future Plans:

  • Potential B2B offering for legal and healthcare sectors where local LLMs address privacy concerns
  • Vision to reduce demand for AI data centers by deploying Mac Studio on-premise with substantially less power consumption

The Local AI Vision

Pae believes local AI's resource requirements will decrease over time:

"The intelligence per wattage — which is the metric for local AI — has been going up significantly. Last year, local AI could barely finish sentences, but today it can actually run tools, write code, access your browser, and order stuff from Amazon … It's just getting better and better."

Key Takeaways

  • Privacy-first architecture: Keep your data on your own hardware
  • Flexibility: Switch between local and cloud models based on task requirements
  • Security: Virtual sandbox isolates AI operations
  • Consumer-friendly: Non-developer-focused interface
  • Open source: Community-driven development
  • Growing ecosystem: 20+ native plug-ins with MCP compatibility