Ollama

Ollama is an open-source platform (MIT License) for running and managing large language models locally on the user's machine or in Ollama Cloud. The project first launched on July 7, 2023 and by July 2026 had grown a community of 9 million active developers, becoming the de facto standard for running open-weight models locally.

Ollama is written in Go and uses llama.cpp as the inference backend for local models. It exposes three surfaces: a command-line interface (ollama pull, ollama run, ollama launch), a native desktop app, and a local REST API (port 11434 by default). Official Python and JavaScript clients let developers plug Ollama into any application.

The model library covers open-weight families such as Llama, Gemma, Mistral, Qwen, DeepSeek, gpt-oss, Phi, plus partner models (NVIDIA Nemotron, MiniMax M3, GLM-5.2). Ollama handles weight downloads, quantization, backend selection (CPU, CUDA, ROCm, Metal, Apple MLX) and runtime — the user needs a single command.

Ollama Cloud, launched in 2025, offers paid access to larger models on datacenter-grade hardware (NVIDIA Blackwell), in US, Europe and Singapore regions, with a zero data retention guarantee. Pricing plans: Free (included with an Ollama account), Pro ($20/month — 3 cloud models running in parallel), and Max ($100/month — 10 cloud models in parallel, 5× more usage than Pro).

The platform integrates natively with agentic tools and coding assistants: Claude Code, OpenAI Codex, OpenClaw, GitHub Copilot CLI, Hermes Agent. In March 2026 Ollama added preview support for Apple MLX. Experimental image generation support (Flux) is available on macOS.

The project is developed by Ollama Inc., headquartered in Palo Alto (California), founded by Jeffrey Morgan and Michael Chiang. In July 2026 the company announced a $65M Series A round, raised alongside a milestone of reaching 9 million active developers.