Ollama

Ollama is an open-source tool for running large language models locally on Mac, Linux, and Windows — and, as of 2026, in the cloud as well. The original single-command workflow (ollama run llama3) still works for pulling and running open models like Llama, Mistral, Phi, Gemma, and Qwen entirely on your own hardware, with a REST API compatible with Open WebUI, Continue, and most local-AI frontends.

Since mid-2025, Ollama has expanded well beyond a CLI tool. It now ships a native desktop app for macOS and Windows with a chat GUI, drag-and-drop file/image support, and a central hub for managing installed models. ollama launch (added June 2026) spins up a full coding agent in one command, with environment variables configured and the model downloaded automatically if missing.

Turbo / Cloud Models

Ollama Turbo runs larger, frontier-scale open models (plus hosted options like GLM, DeepSeek, and Kimi) on Ollama's own datacenter-grade hardware instead of local GPUs, for users whose machines can't run bigger models. Turbo is a paid tier layered on top of the still-free core product: Free (light cloud usage, 1 concurrent cloud model), Pro ($20/month, larger models, 3 concurrent, ~50x more usage than Free), Max ($100/month, currently paused for new signups), Team ($25/seat/month, zero data retention), and custom Enterprise pricing. Local model execution on your own hardware remains unlimited and free on every tier.

Ollama remains the default backbone for the local-AI developer ecosystem, now bridging local and cloud inference under one interface.