Ollama
Ollama is an open-source tool under MIT for running open models on your own computer. One command downloads a model from its library and starts it, and a REST API on your machine lets other programs use it. It is built on llama.cpp, has official Python and JavaScript libraries, installs on macOS, Windows, Linux and Docker, and can launch coding tools such as Claude Code, Codex and OpenCode against a local model.
What does Ollama do?
Ollama's README opens with "Start building with open models." You install it, run a model by name, such as ollama run gemma4, and it downloads the model from the library at ollama.com/library and starts it. "Ollama has a REST API for running and managing models", served on your machine at port 11434, and official libraries for Python and JavaScript wrap it. Its backend is the llama.cpp project, which the README credits by name.
It also connects agents to local models. The README says you will be prompted to run a model "or connect Ollama to your existing agents or applications such as Claude Code, OpenClaw, OpenCode, Codex, Copilot, and more", and a launch command starts those tools against a model Ollama is serving. Models can be imported and customised with a Modelfile. The README itself does not list hardware support, model formats, or an OpenAI-compatible endpoint; those are in the linked documentation, which this page did not read, so they are not stated here.
Key facts
- Licence: MIT, copyright Ollama, with no added conditions.
- One command to download and run a model from the ollama.com library, for example ollama run gemma4.
- A REST API for running and managing models, at localhost:11434 in the README's example.
- Backend: llama.cpp, credited in the README.
- Official client libraries: ollama-python (pip install ollama) and ollama-js (npm i ollama).
- A launch command that connects coding agents such as Claude Code, Codex, OpenCode and Copilot CLI to a local model.
- Install: a shell script for macOS and Linux, a PowerShell script for Windows, the macOS and Windows downloads, or the ollama/ollama Docker image; community packages for Homebrew, Nix, Helm and others.
- Latest tagged release when read: v0.34.2 on 2026-09-15; 181,180 stars on 2026-09-17, the most of any project in this directory after OpenCode.
What does it replace, and where does it fall short?
Ollama replaces calls to the OpenAI API, for work that an open model can do, with a model running on your own machine behind a local API: nothing is sent out and nothing is billed per token. It is the simplest way on this site to get a local model running, and most of the chat interfaces and agents in this directory list it as a backend, which is why it leads the open-source alternatives to the OpenAI API page.
Where it falls short: its README does not describe OpenAI API compatibility, hardware support or multi-GPU serving, so code written against the OpenAI SDK should be checked against its documentation before switching. It is built for one machine; for high-throughput serving to many users, vLLM. The quality of the answers is the quality of the open model you choose.
How does Ollama run?
On macOS or Linux run the install script from ollama.com, on Windows the PowerShell script, or download the installer; or run the ollama/ollama Docker image. Then ollama run and a model name. The README lists no system requirements. It describes no hosted or paid Ollama service; its cloud section links third-party deployment guides.
Who is Ollama for?
Anyone who wants an open model running locally in a minute, for a chat app, a coding agent or their own scripts. A team serving many users from GPUs should look at vLLM; someone who needs images, audio and embeddings from one server should look at LocalAI.
What limits does the README state?
None stated in the README.
Questions people ask
Is Ollama open source?
Yes. Ollama is MIT licensed with no added conditions. It builds on llama.cpp, which is also MIT. The README describes no paid tier from Ollama; its cloud section links only third-party guides for deploying it yourself.
Is Ollama compatible with the OpenAI API?
Its README does not say. The README describes Ollama's own REST API for running and managing models, at port 11434, and links further API documentation that this page did not read. Several apps on this site call Ollama through an OpenAI-compatible bridge, so check the docs for the endpoint your code needs.
Which models can Ollama run?
Open models from its library at ollama.com/library, downloaded by name; the README's example is Gemma 4. You can also import models and customise them with a Modelfile, both covered in linked documentation.
How does Ollama compare with the OpenAI API?
Ollama runs open models on your own machine for free, with nothing leaving it, behind a local API. The OpenAI API gives OpenAI's models as a hosted service billed by use. The open-source alternatives to the OpenAI API page compares Ollama with LocalAI, llama.cpp, vLLM and the desktop apps that serve an API.
Can Ollama run coding agents like Claude Code?
Yes, by its README: it can connect a local model to Claude Code, Codex, OpenCode, Copilot and others, and has a launch command for them. The agent then uses the model Ollama serves rather than its vendor's hosted model.
Sources
- Ollama README and MIT LICENSE: github.com/ollama/ollama, read 2026-09-17.
Compared with the others
On the open-source alternatives to the OpenAI API page, Ollama is the pick for the quickest local start. MIT, one command downloads and runs an open model behind a local REST API, built on llama.cpp, with Python and JavaScript libraries and a launcher for coding agents.
Also on that page: LocalAI for every modality behind openai-compatible apis, llama.cpp for the engine on any hardware, vLLM for serving many users from gpus.
More local model runtimes
The MIT C and C++ inference engine behind much of local AI, running GGUF models with 1.5 to 8-bit quantisation on CPUs, Apple Silicon and 17 hardware backends, with an OpenAI-compatible server and a web interface built in.
An Apache-2.0 engine from UC Berkeley's Sky Computing Lab for high-throughput model serving, with an OpenAI-compatible and Anthropic Messages API, 200+ model architectures, wide quantisation support, and parallelism across many GPUs.
An MIT engine that runs text, vision, voice, image and video models on any hardware, no GPU required, behind drop-in OpenAI, Anthropic and ElevenLabs-compatible APIs, over 60+ backends including llama.cpp, vLLM and whisper.cpp.
Added September 17, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 17, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/ollama/ollama.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.