Local Models

Local models run on your own machine, giving you more control over privacy, performance, and offline workflows.

Msty Studio provides one unified Local Models area for models running through Ollama, MLX, and Llama.cpp.

Open Local Models

Open Model Hub > Local Models.

The Local Models area has four tabs:

• Featured
• Installed
• Model Search
• Import

Featured

Featured provides a curated catalog of models with available Ollama, GGUF/Llama.cpp, and MLX formats grouped together.

• Use the engine and format filters to narrow the catalog
• Review model purpose, size, and hardware-fit information
• Install using the recommended engine or choose another available format
• If the selected engine is not installed, Msty sets up the engine before installing the model

On Apple Silicon, Msty generally recommends MLX when an MLX format is available. Recommendations on other systems depend on the available hardware and supported engines.

If you are unsure which model best fits your use case, use Model Matchmaker.

Installed

Installed combines models from all managed local engines into one list.

• Filter or browse models without switching between engine-specific pages
• Review the engine, model location, and configuration
• Rename a model or update its purpose tags
• Edit supported model configuration and chat-template settings
• Uninstall a model

Model Search

Model Search combines local model discovery into one workflow.

• Search the Ollama registry and Hugging Face
• Filter results by Ollama, GGUF, or MLX format
• Install directly to a compatible engine
• Choose between Ollama and Llama.cpp when a GGUF model supports both paths

Import

Use Import for models already stored on your computer.

Import a GGUF file

1. Open Model Hub > Local Models > Import.
2. Select Select GGUF File.
3. Choose the .gguf file.
4. Msty imports it into an available compatible engine or guides you through engine setup.

Import a Safetensors or MLX directory

1. Open Model Hub > Local Models > Import.
2. Select Select Directory.
3. Choose the model directory containing its Safetensors or MLX files.
4. Msty detects the model format and uses an available compatible engine.

Installation Progress

Model and service installations continue while you use other areas of Msty Studio.

• The Model Hub sidebar badge shows the number of active installs
• The active installs panel shows download progress
• Downloads can be cancelled from the active installs panel
• Completed models appear in the Installed tab

Choosing an Engine

Ollama

Use Ollama for broad model availability and a straightforward local inference workflow. Msty Studio's managed Ollama integration is labeled Local AI in service settings.

MLX

Use MLX on Apple Silicon for models optimized for Apple's unified-memory hardware.

Llama.cpp

Use Llama.cpp for GGUF models, flexible hardware support, and deeper runtime controls.

Manage Local Services

Local engine service controls remain available in Settings:

• Settings > Local AI
• Settings > MLX Service
• Settings > Llama.cpp Service

Use these pages to review service health, endpoints, versions, logs, start/stop controls, and supported acceleration options.

Llama.cpp Model Parameters

When a Llama.cpp model is selected in chat, open Model Parameters for Llama.cpp-specific settings:

• Num ctx controls the runtime context window
• Truncate Middle keeps early and recent context while trimming the middle
• Truncate Old removes the oldest messages first
• None disables automatic history trimming and may exceed the model context limit

General GPU Support

If your system has a compatible GPU, enable supported acceleration from the relevant service settings.