Hardware-aware · Open source · Updated weekly
The hub for local LLMs, install guides and a hardware picker.
You can run AI models like the ones behind ChatGPT or Claude directly on your laptop or desktop, without sending your data anywhere. RunLocal shows you which open source model to choose for your hardware, which free software to install (Ollama, LM Studio, llama.cpp), and how to get started in about ten minutes. No prior knowledge required.
New to this? Start with the glossary for plain-language definitions, or jump straight to the hardware picker to see what your computer can run.
A short list, opinionated. Full catalog in the directory.
Alibaba (Qwen team) · China
Current open-weight Qwen generation for local use. Qwen3.6-27B is the key workstation-class checkpoint. Qwen3.8-Max-Preview is newer but cloud/API-only as of August 15, 2026 and is intentionally excluded from the local picker.
Alibaba (Qwen team) · China
Released in 2026. The smaller checkpoints remain excellent local choices when Qwen 3.6 is too large.
Google DeepMind · United States
Multimodal family spanning edge through workstation hardware. The 26B-A4B model activates about 4B parameters per token while retaining a 26B memory footprint.
Mistral AI · France (EU)
Combines instruct, reasoning and coding modes in one multimodal MoE model. Low active parameters help compute speed, but the full weights still determine memory use.
Snapshot · Aug 17, 2026
Auto-curated from the Hugging Face Hub by a weighted mix of downloads, community likes and recency. Refreshes weekly via GitHub Action. License tier is a quick visual hint, not legal advice.
deepseek-ai
MIT
zai-org
MIT
deepseek-ai
MIT
ornith-ai · 35B
MIT
microsoft
MIT
deepseek-ai
MIT
zai-org
MIT
prism-ml · 27B
Apache 2.0
Runtimes, GUIs and inference servers, with their real trade-offs.
Runtime
The fastest way to get a local LLM running with one command.
Runtime
Maximum control and the broadest hardware coverage in the open ecosystem.
GUI
Browsing, comparing and chatting with local models in a desktop GUI.
Server
Production-grade inference with concurrent users and high throughput targets.
From zero to a working local LLM in about ten minutes, with the commands that actually matter and the gotchas nobody warns you about.
How to compile llama.cpp with the right backend for your hardware, pick a GGUF quantization that fits your RAM, and serve an OpenAI-compatible endpoint.
A practical walkthrough for using LM Studio to download, compare and serve local models, plus when LM Studio is the wrong tool.
A 2.8-trillion-parameter model with public weights that almost nobody can run. What the largest open weight ever released actually means for people who run AI locally.
Q4, Q5, Q8 and the rest of the GGUF zoo, with a practical decision rule that holds up across hardware classes.
Unified memory, raw VRAM, and the workloads where each wins. A practical comparison that goes beyond benchmark snippets.
Rolling releases, immutable variants, and an honest line between community Linux and paid enterprise. A practical look at when openSUSE earns its place in an AI stack.
Five frontier-class releases in the last thirty days, three of them from Chinese labs. A short tour of where the field actually is.
Ollama, llama.cpp, LM Studio and vLLM solve different problems. A practical map of when to reach for which, and why it matters more than benchmark numbers.
The big cloud services are easier to start with. But there are real reasons to do it yourself. Three of them.
What you type and what the model answers never leave your computer. Handy when you are working with personal notes, client documents, internal code, or anything you would not paste into a public website.
The model file lives on your disk. If the company that made it shuts down, raises prices, or simply changes its terms, your setup keeps working. The model you download today still runs in 2030 if your computer does.
Cloud AI charges per use. Local AI costs you the price of your computer, plus electricity. After the first month, the marginal cost of an extra question is essentially zero.
