Ollama is an open-source software platform developed by Jeffrey Morgan and Michael Chiang in 2023 for running and managing large language models on local computers and through hosted cloud models. It provides a command-line interface, a native GUI, a local REST API, model-management tools, and integrations for using open-weight models with coding assistants and other applications. The platform is designed to simplify the process of downloading, running, and managing models such as Llama, Gemma, Mistral, Qwen, and DeepSeek on personal hardware, making generative AI more accessible to individual users and developers.
Ollama operates as a bridge between the raw model weights and the end user, abstracting away the complexities of model serving and inference. By default, it runs locally, which addresses privacy and data-control concerns that arise with cloud-based AI services. The project has gained traction within the open-source community as a tool for experimentation, development, and deployment of open-weight models without requiring extensive infrastructure or cloud subscriptions.
History
Ollama was first released on 8 July 2023. The project became associated with the growth of local large language model software, allowing users to download and run models such as Llama, Gemma, Mistral, Qwen, gpt-oss, GLM, and DeepSeek on a local machine. This release coincided with a broader trend toward local inference, driven by improvements in model quantization and consumer hardware capabilities.
In 2025 and 2026, Ollama added additional application and cloud features, including hosted cloud models, web search support, tool and coding-agent integrations, and support for using Ollama with applications such as Claude Code, Codex, OpenCode, Copilot CLI, and OpenClaw. In March 2026, Ollama announced preview support for Apple silicon, expanding its compatibility with Mac hardware. In July 2026, the company raised $65 million in funding, signaling investor confidence in the platform's growth and market position.
Features
Ollama includes tools for downloading, running, importing, and managing large language models. Users can run models from the command line, interact with them through a local HTTP API, or use client libraries for programming languages such as Python and JavaScript. The project provides a REST API for chat and model-management functions, with the default local service commonly exposed on port 11434. Ollama also distributes an official Docker image and provides model libraries and documentation for running supported models.
The platform uses the llama.cpp backend for local model inference, which supports running of quantized models that use less memory. It can employ graphics cards (GPUs) to speed up calculations, leveraging hardware from vendors such as NVIDIA, AMD, and Intel. Ollama supports a model-library format that allows users to pull, run, and manage model variants by name, simplifying the workflow for switching between different models or configurations.
Meanwhile, the project's downloadable software is being explored as a back-end platform for low-code/no-code self-hosting of LLMs, potentially allowing natural language processing use cases to become more accessible to nontechnical end users. Ollama also has experimental support for image generation, using models such as Flux, on macOS, with Windows and Linux support expected in the future.
Security
Because Ollama is commonly used to run local or self-hosted AI models, security researchers have examined risks from misconfigured public deployments. In January 2026, The Hacker News reported on research by SentinelOne and Censys that found many Ollama servers were exposed to the public internet, mainly by binding to the INADDR_ANY interface (0.0.0.0), which on an inadequately-secured system would allow access from other local or remote devices despite the fact that Ollama is intended to run locally by default. This highlights the importance of proper network configuration and firewall rules for users deploying Ollama in shared or cloud environments.
Ecosystem and Integrations
Ollama's design as a local inference server has made it a popular backend for various applications. It integrates with coding assistants like Copilot and other developer tools, enabling users to run open-weight models for code completion and generation without sending code to external servers. The platform also supports web search features, allowing models to retrieve up-to-date information during conversations, and tool integrations that enable models to call external functions or APIs.
The model library maintained by Ollama includes a wide range of open-weight models from organizations such as Meta (Llama), Google (Gemma), Mistral AI (Mistral), and Alibaba Cloud (Qwen). This diversity allows users to choose models based on performance, size, or licensing requirements. Ollama's command-line interface and REST API make it suitable for scripting and automation, and its Docker image facilitates deployment in containerized environments.