Skip to content
← Tools / Skills
ToolOllama Inc./10 trustVerified
Ollama logo

Ollama

Run open-weight large language models locally with a CLI and OpenAI-compatible API.

Visit official site →

About

Ollama runs open-weight large language models locally on macOS, Windows, and Linux, exposing them through a CLI and an OpenAI-compatible REST API on localhost:11434.

Models are packaged as Modelfiles and pulled from the ollama.com registry, with automatic GPU offloading for NVIDIA, AMD, and Apple Silicon hardware.

It ships official Python and JavaScript SDKs, is MIT-licensed, and had roughly 183k GitHub stars as of October 2026. In 2026 Ollama also added sign-in cloud models, so the local-only privacy story applies to local model tags.

What it can do

  • Runs open-weight LLMs locally with a simple `ollama run <model>` CLI
  • REST API on localhost:11434 plus OpenAI Chat Completions / Responses and Anthropic Messages compatibility
  • Modelfile format for customizing, importing (GGUF, safetensors), and sharing models
  • Automatic GPU acceleration (NVIDIA CUDA, AMD ROCm, Apple Silicon Metal) with CPU fallback
  • Hundreds of models in the ollama.com/library registry (Llama, Gemma, Qwen, DeepSeek, gpt-oss, and more)
  • Vision-capable models accept images alongside text prompts
  • First-party Python (ollama-python) and JavaScript (ollama-js) SDKs
  • Direct Hugging Face pulls via `ollama run hf.co/user/repo:quant`
  • Launch integrations for coding agents (Claude Code, Codex CLI, OpenCode, and others)

Setup

  1. Download and install Ollama for your platform from ollama.com: on macOS or Linux run curl -fsSL https://ollama.com/install.sh | sh; on Windows run irm https://ollama.com/install.ps1 | iex. A Docker image (ollama/ollama) is also available on Docker Hub.
  2. Start the Ollama app. On Linux, start the server with ollama serve if it is not already running.
  3. Pull and chat with a model: run ollama run gemma4 — Ollama downloads the model and opens a chat in your terminal (type /bye to leave). Browse the full model list at ollama.com/library.
  4. Optional: connect a coding agent from your project directory, e.g. ollama launch claude to launch a supported integration.
  5. Optional: use the API — send requests to http://localhost:11434 (REST or OpenAI-compatible endpoints). Install an SDK with pip install ollama or npm i ollama.

Prerequisites

macOS, Windows, or Linux (8 GB RAM minimum; GPU recommended for larger models)

Use cases

chatcodevisionreasoningdata

Works with

Generic APIAny agent (.md)

Trust score — 9/10

/10 trustVerified

Extremely active (commits Oct 12, 2026), ~183k stars, permissive MIT license, no dead-project risk. The one caveat: in 2026 Ollama added sign-in cloud models alongside local ones, so the "everything stays local" privacy story now applies only to local model tags — docs disclose this, and inference itself remains local after download. Metrics-v1 score 9/10 — adoption 9.8, maintenance 9.1, openness 10, manual 9.

★ 182,876 GitHub stars

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
9.8/10
Maintenance · recency, releases, issue load
9.1/10
Openness · license + docs
10.0/10
Manual review · sanity read + red-flag check
9.0/10
  • · 182,876 GitHub stars (log-scaled vs catalog peers)
  • · blended with ~13,275,724 monthly package downloads
  • · last push 0d ago; 100 releases/12mo; 4248 open issues
  • · manual review: no red flags; 2026 cloud-models caveat to local-only privacy story

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.