Ollama’s new app just made running local LLMs super easy

Ollama, an open-source platform for running large language models (LLMs) locally on a user’s machine, has released a new macOS and Windows app that allows users to easily download and run AI assistants.

The new app has a minimalistic graphical interface that allows users to easily download and chat with LLMs.

It supports file dragging and dropping, enabling simple reasoning with text or PDF documents.

Users can select the specific context lengths with a slider, determining how many conversations local LLMs can remember and use to generate responses. For large documents, the context length sliders go up to 128k tokens.

Ollama’s app also includes multimodal support, meaning that LLMs, such as Google DeepMind’s Gemma 3 model, can analyze images and process code files.

However, the app lacks more advanced features currently available on proprietary AI platforms, such as previews of generated code or designs, or image generation.

The app's LLM selection currently seems to be limited to small Gemma, Deepseek, and Qwen options.

Other tools, such as LMStudio, are available for running LLMs locally. However, Ollama believes that its app has the advantages of quality, ease of use, model support directly with the major labs, and hardware support. However, some users noted that the macOS Ollama version does not support MLX, a native machine learning framework optimized for Apple silicon.

Until recently, Ollama was available as a command-line interface (CLI) tool for more advanced users. The platform continues to maintain the versions for developers and other power users.

Ollama’s statistics reveal that the platform users have pulled popular open-source models hundreds of millions of times. DeepSeek R1 variants have been downloaded 55.2 million times, Llama 3.1 99 million times, Nomic embed text 34.5 million times, and Llama 3.2 27.5 million times. Many more models from Google, Mistral, Microsoft or Alibaba Cloud also have millions of pulls.