Run AI on your own machine. Pick your hardware, install one tool, chat with a model that fits — this page gets a beginner there in about 15 minutes.
Backstory · Follow updates on X
Localist is a beginner-first guide, not another giant tool list. You start from the machine you already own, follow one guide, and skip the other 50 options. And it doesn't rot: a pipeline — not vibes — refreshes the Fresh updates section every morning, and we review and prune stale recommendations instead of hoarding them.
Jump to: Start here · The essentials · Fresh updates · Glossary · Contributing
Not sure what hardware you have? → Work it out in two questions · Something not working? → Troubleshooting
Don't read everything. Find your row, follow one guide, and you'll be chatting with a local model in under 15 minutes.
| I have… | Your guide |
|---|---|
| An NVIDIA GPU (any GeForce RTX) | NVIDIA GPU path |
| A Mac with Apple Silicon (M-series) | Mac path |
| An AMD GPU (Radeon RX) | AMD GPU path |
| Just a laptop / no GPU | CPU-only path |
| A mini PC or home server | CPU-only path — or the NVIDIA path if it has a GPU |
| No idea what I have | Start here |
Everything here can run locally; many paths work offline after setup, and your prompts stay on your machine unless you connect a cloud service.
Stuck partway? Troubleshooting covers the handful of things that actually go wrong — including "where do I type this command?", how to stop the chat, and how to start it again tomorrow.
One opinionated pick per category. Alternatives are inside each entry — but if you're new, just take the pick and move on.
Which tool first? Want clicks? Start with LM Studio. Comfortable with one command? Start with Ollama. You do not need both.
| Category | Our pick | Why this one |
|---|---|---|
| Model runner | Ollama | One command to install, one to run a model. The de-facto beginner standard. |
| Desktop app | LM Studio | Point-and-click everything: browse, download, and chat with models. No terminal needed. |
| Web UI | Open WebUI | ChatGPT-style interface on top of Ollama. Multi-user, chat with your documents, voice. |
| Chat models | Qwen3 family | Strong at every size, tiny to huge. Which size fits you → choosing models. |
| Coding models | Qwen Coder | Same idea, meaningfully better at code. Pairs with the coding agent below. |
| Coding agent | OpenCode | Most popular open-source coding agent; points at your local models. |
| Image generation | ComfyUI | Node-based, runs every major open image model. Steeper curve, unmatched power. |
| Chat with your docs | AnythingLLM | Point it at a folder and a local model — done. |
| Speech-to-text | whisper.cpp | Fast local transcription on any hardware. |
| Text-to-speech | Piper | Fast, natural offline voices — runs even on a Raspberry Pi. |
When you outgrow beginner tools, llama.cpp is the engine worth learning — it's what most of the tools above are built on.
The full curated set (with licenses, hardware fit, and honest caveats) lives in
data/curated.yml — it's the single source of truth these
picks come from.
New projects and tool releases from the past week, refreshed daily by the pipeline. Pipeline-surfaced projects are not curated recommendations. Start with the Essentials table above if you want the trusted beginner picks.
Updated 2026-11-01
🆕 New & active projects
- peonist-ai/halogen-flash-server — Local inference that behaves like a frontier API: Qwen3.8-Flash-Next on AMD Strix Halo plus small NPU models, over the OpenAI and Anthropic APIs on one port. · ⭐ 1130
- Nomily-Ai/nomily-app — BYOK voice recording companion with ASR transcription, speaker diarization, AI summaries, and local ASR endpoint support. · ⭐ 1018
- AienTech/dossier — A self-hosted job search app that writes tailored resumes and cover letters from your profile. · ⭐ 533
- prabanta-dev/janas — 37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs 16 tok/s. Faster than llama.cpp on all 13 models compared in Q4_K_M. LLMs bigger than · ⭐ 533
- dragonked2/alphacode — Free MIT AI coding agent — no API key needed. Built-in free model, or bring Claude, GPT, Gemini, DeepSeek, Ollama +50 more. Swarm mode, 40+ tools, browser & desktop automation, built in Rust. · ⭐ 516
📦 Tool releases
- ComfyUI v0.43.0 — model_prefetch: syncrhonize all graph cleanups Add rope scaling support to MOSS-image
- Ollama v0.40.12 — Liquid AI LFM3 [LFM3]( is Liquid AI's new family of small models built for local agents, tool calling, and structured output, on Liquid's new tri-hybrid architecture. The family includes:
- Jan v0.8.8 — Jan v0.8.8: security fixes, llama.cpp 0.7.0, and Trusted Hosts for the local API A security update to v0.8.7. We recommend updating. Your data and settings carry over; a few behaviours change, listed
- Ollama v0.40.13 — LFM3 on Linux and Windows Liquid AI's [LFM3]( models, including text-to-speech in [lfm3-audio]( and [lfm3-omni]( now run on Linux and Windows as well as Apple silicon.
- Open WebUI v0.12.5 — Added 🧭 Analytics session explorer. Administrators can now open a Sessions tab in Admin Analytics that lists each chat with its tokens, responses, errors and time span, filters them by search,
- LM Studio 1.1.10
- KoboldCpp v1.124 —
[ a e s t h e t i c . e d i t i o n ]NEW! - KoboldCpp has now integrated a small subset of popular models from [audio.cpp]( by @0xShug0! This greatly enhances the available built-in Music Gen - KoboldCpp v1.124.1 —
[ a e s t h e t i c . e d i t i o n ]NEW! - KoboldCpp has now integrated a small subset of popular models from [audio.cpp]( by @0xShug0! This greatly enhances the available built-in Music Gen
Don't know what hardware you have? Start here works it out in two questions, and tells you what a local model will and won't do well before you spend time on it.
The glossary explains every term you'll bump into — GGUF, quantization, weights, context window, VRAM, tokens/sec — in plain words. And choosing models answers the #1 question: which model size actually fits my machine?
When something doesn't work, troubleshooting is organised by what you see on screen — not by what went wrong underneath.
Found a great tool? Spotted a dead project? Open an issue — suggest a tool · report a stale entry. If a link, download, or command looks unsafe or compromised, report it privately through SECURITY.md instead of a public issue. See CONTRIBUTING.md for the ground rules.