Ollama is an open-source framework designed for running large language models (LLMs) locally on macOS, Linux, and Windows. It enables developers to deploy and interact with advanced AI models through a simplified command-line interface and a local REST API.

• Packages model weights, configurations, and data into a single Modelfile for reproducible local deployments.

• Provides built-in support for diverse architectures including Llama 3.2, Mistral, Gemma 3, Phi-4, and DeepSeek-V3.

• Optimizes hardware utilization by automatically leveraging NVIDIA GPUs or Apple Silicon for accelerated inference.

• Exposes a local server endpoint that integrates with existing development workflows via standard HTTP requests.

• Built with Go (Golang) for efficient system-level performance and cross-platform compatibility.

• Supports quantized model formats to reduce memory overhead while maintaining response quality.

• Develops privacy-focused AI applications that process sensitive data without internet connectivity.

• Benchmarks and compares different LLM variants locally before production deployment.

Download the installer or run the ollama run command to pull and execute your first model.