User Guide
How to Use iRun Studio
Master on-device artificial intelligence. Learn how to browse models, load GGUF/MLX weights, configure inference modes, and deploy autonomous tools.
Download Free on App StoreInstallation & Device Setup
Download iRun Studio directly from the Apple App Store on macOS (Sonoma/Sequoia), iOS 17+, or iPadOS 17+. The app requires zero accounts or registration to begin using local intelligence.
System Tip: For Mac, 16GB+ Unified Memory is recommended to run 9B–14B parameter models with maximum speed. 8GB Macs and iPhones smoothly run 4B–8B Q4 quantized models.
Exploring the Model Browser
Open the right sidebar to access the Model Browser. Models are organized into three clear categories:
- Language: General chat, coding, and multi-turn reasoning (Gemma, Qwen, DeepSeek, Llama, Bonsai).
- Vision: Multimodal models for image inspection and visual Q&A.
- Logic: Specialized mathematical and structured thinking models.
You can download models with 1 click via the built-in Hugging Face browser or manually drop .gguf files into Finder > iRun Docs > LLM-models.
Choosing Inference Modes
Tailor how the AI thinks and responds directly from the chat input pills:
Optimized for instantaneous answers, translations, and short coding snippets.
Allocates extra chain-of-thought tokens before streaming the answer.
Enables autonomous tool calling, deep context memory, and subagent delegation.
Autonomous Tool Execution
When asked for up-to-date facts or complex tasks, iRun automatically invokes tools (such as web_search or terminal inspection). You can monitor tool execution in real-time with live timer badges and expand citation cards directly in the chat feed.
CoreML Diffusion & Workstations
Switch between tabs in iRun Studio to generate images with CoreML Stable Diffusion, build visual IdeaGen concepts, or run multi-file research tasks with DeepSearch.