Open source & local AI
Models and apps you can run yourself. The license is part of the verdict.
A best-for note is a starting point for a stated job. It is not a ranked bake-off. We did not run the products, and your laptop does not owe us a benchmark.
Best for
UseRun a chat model on your computerOllama
The short path from a model name to a local server. Your laptop fan will have opinions.
Read the profile →
UseServe a model to other softwarevLLM
An Apache-licensed server that lets apps talk to your model the way they talk to OpenAI, on hardware you control.
Read the profile →
WatchA desktop app with a windowLM Studio
Free to download, and not open source. Read the terms before you hand the installer to the whole company.
Read the profile →Tool verdicts
WatchComfyUI
A node graph for local image and video models.
Free under GPL-3.0. Nodes and models you add can carry other licenses.
UseCrewAI
A Python framework for role-shaped agents.
The library is MIT. Model calls are billed by the provider you choose.
UseFlowise
A visual builder for LLM flows you host yourself.
Free under Apache 2.0. Any model you connect is billed separately.
WatchJan
A desktop app that runs local models and exposes a local API.
Free to download. The README states Apache 2.0. Confirm the LICENSE file you install.
Usellama.cpp
A local inference engine for GGUF-style model files.
The engine is free under the repository license. You supply the model.
WatchLM Studio
A desktop app for chatting with local models.
Free to download. The desktop app is not open source.
Watchn8n
A workflow tool with AI nodes. Source-available, not OSI open source.
Starter is 20 euros a month billed annually. Pro is 50 euros. The community edition is free for your own internal use.
UseOllama
A local runner for open-weight chat models.
Free to download. The app license is MIT. Each model you pull has its own license.
WatchOpen WebUI
A chat interface you host in front of Ollama or another model server.
Free to self-host. The license restricts branding changes above a stated user count.
UsevLLM
An Apache-licensed server for open-weight models.
Free under Apache 2.0. You pay for the GPUs.
Resources
Places to get weights, read a license, or look at someone else's leaderboard. A number on those sites belongs to them.
Check a bill
- AI subscription and usage-limit checker. Public plan prices and limits, with the day we checked. Overlap and a cheaper sticker. The email box stores nothing until beehiiv is connected.
Model hubs
- Hugging Face models. Weights, plus the model card that states the license. Read that card before the group chat does.
- Ollama library. Models packaged so you can pull them into a local Ollama server. Your fan will notice.
- Llama downloads. Meta's download page for Llama weights and the community license. The license is the homework.
Leaderboards
- LMSYS Chatbot Arena. A public preference board. Their rank is not a Loaded Tokens verdict, even if the screenshot looks official.
- Artificial Analysis. Independent model comparisons. Read their method before you paste a number into Slack.
- Open LLM Leaderboard. Hugging Face's board for open models. The score stays on their page. We did not adopt it.
Learning
- Loaded Tokens learn tracks. Five tracks on this site. The diagrams are sketches of an idea, not benchmarks.
- Hugging Face course. The hub's own lessons on models, datasets, and inference. Their classroom, their syllabus.
Local runtimes
- llama.cpp. The local engine. The model you load still has its own license, like a roommate with a separate lease.
- vLLM docs. How to serve an open-weight model to software you run. The docs are the instruction manual, not a verdict.
- Whisper. OpenAI's speech model for local transcription. This download is not the hosted API.
All resources · Full directory · This category as JSON
Updates that matter in five minutes or less.
Free, Tuesdays & Fridays.