The practical guide to open-source AI

Check your PC. Choose your models. Build your local AI stack.

Six free tools take you from “what can my machine run?” to a working local setup you can verify. No account, no paywall.

What are you running?

Don't know your specs?

Free, no signup. Your specs never leave your browser.

Plan a new AI PC build →New? Start Here →

98 tools reviewed25 hardware presetsFree — no signupSpecs never leave your browser

Step 0 · Orientation

How to use OpenSourcesAI

SYSTEM BRIEF: HOW TO USE OPENSOURCESAI

OpenSourcesAI is a decentralized, privacy-first technical workspace built to help you escape cloud vendor lock-in. Instead of paying recurring API subscription fees, use this interactive roadmap to audit your local graphics hardware (or plan a new build if you're buying), isolate compatible model versions (unquantized for power users, or GGUF—highly compressed files that fit regular laptops), and copy clean, ready-to-run startup scripts (zero dependencies means they work instantly without installing complex code libraries) to run open-weight models locally on your own terms.

$ ollama run qwen3:8b

CHECK

Audit the PC you have. The Compatibility Checker matches your VRAM and RAM to models that actually fit.

PLAN

Buying or upgrading? The PC Builder turns your target models and budget into three build tiers.

CONFIGURE

The Model Wizard picks a model, runtime, and quantization for your workload — with copy-paste setup.

TEST

Load a small model in the browser Playground and feel how prompts behave before you commit.

Most people start here

Local LLM Compatibility Checker

Pick a hardware preset — or enter your GPU, VRAM, and RAM — and get realistic local model matches with run commands, right on this page. No signup, and your specs never leave your browser session.

Optional helper

Auto-detect my device basics

We ask your browser for approximate memory, CPU thread count, and graphics renderer info. This is not a full PC scan — review and correct any field before running the checker.

Local LLM Hardware Compatibility Matrix

Select your graphics card memory (VRAM) and system RAM. The grading engine matches your hardware against each model’s minimum requirements. We require at least 2 GB of VRAM headroom above the model’s minimum before rating a fit as Comfortable.

Your hardware

Quick-select a hardware presetPresets are optional shortcuts covering CPU-only machines, NVIDIA gaming cards, Apple Silicon, and multi-GPU workstations (the 2× and 4× GPU rigs at the end of the list). Selecting one fills VRAM, RAM, and GPU name instantly — adjust any field to fine-tune. Card not listed? Skip the presets and set VRAM and RAM directly below.

CPU — optionalOptional — helps the wizard generate better runtime notes. Does not change the match.GPU — optionalOptional context — the match is graded from VRAM and RAM. The GPU name refines Apple Silicon unified-memory handling and deployment notes.GPU VRAM in GB — requiredSelect 0 for integrated GPUs or CPU-only builds. For Apple Silicon, use your unified memory figure. Multiple GPUs? Select the combined total across cards — 2× RTX 3090 is 48 GB. Exact size not listed? Round down.

What is VRAM?

VRAM is your graphics card's dedicated memory. Larger models need more VRAM — a 70B parameter model needs roughly 42–47 GB, while an 8B model typically needs 6–8 GB. Quantization (like Q4_K_M) shrinks the model, letting smaller cards run bigger models. Learn more about VRAM →

Two GPUs, a laptop GPU, or an unusual VRAM size? How to enter it

Multiple graphics cards: select the combined VRAM total — 2× RTX 3090 is 48 GB, 4× RTX 3090 is 96 GB. Results grade against the combined pool, because runtimes like Ollama split a model across cards automatically; a note about multi-GPU speed appears with your results. The preset list also has ready-made 2× and 4× GPU entries under “Multi-GPU Workstations and Clusters.”
Using auto-detect with two cards: browsers can only see one graphics adapter, so auto-detect reports a single card — raise the VRAM figure to your combined total yourself before running the check.
Apple Silicon (M1–M4): enter your total unified memory as the VRAM figure. The checker automatically sizes against about 75% of it, since macOS shares that memory with everything else you have open.
Laptop GPU: use the card's dedicated VRAM — a laptop RTX 4060 is 8 GB. Ignore “shared graphics memory” in Windows settings; that is system RAM, which the checker already counts separately.
VRAM size not in the list: pick the next size down — an 11 GB GTX 1080 Ti checks as 10 GB. Underestimating keeps the verdict honest.
External GPU (eGPU): enter its VRAM normally. The enclosure slows model loading, not whether a model fits.

Planning a dual-GPU build? Read the dual RTX 3090 guide →

System RAM in GB — requiredIf a model is too big for your graphics card, your computer can run part of it in ordinary system memory instead — slower, but it still works. More RAM keeps more models within reach.Primary workflowWhat you want the model to do — the biggest factor in how results are ranked.

Which workflow should I pick?

Pick the task you will do most often — it carries more weight than any other answer here. Choosing coding moves models tuned for code to the top; choosing document Q&A favours models that handle long documents and retrieval well. Nothing is locked in: change it and run the check again to compare. Compare models by use case →

PreferenceNudges the ranking toward faster or higher-quality versions of each model.

Which preference should I choose?

The same model comes in different sizes on disk. Speed first favours the more compressed versions — they load quicker and leave VRAM to spare, with a small drop in answer quality. Quality first favours the fuller versions, which answer better but need more memory. Balanced sits between the two, and is the right default if you are unsure. How model compression works →

Scores use practical VRAM fit rules — not live benchmarks. Confirm specs against your GPU manufacturer before long inference sessions.

Results are ranked for: Chat and general assistant · Balanced quality and speed — change the two fields above to match your goal.

Set your GPU VRAM or system RAM above to run the check — at least one of the two is needed. Not sure? Use auto-detect or pick a preset.

Built by OpenSourcesAI · Free · No signup

Start with the right tool

These six are ours — built and maintained by OpenSourcesAI, free to use, and designed to hand off to each other. New here? Get a guided path on the Start Here page →

Already have a PC?

Compatibility Checker

See which local models realistically fit your GPU, VRAM, and RAM — with run commands included.

Also on this page — jump to the checker ↑

~60 seconds · no signup

Open the full checker →

Planning a build or upgrade?

Local AI PC Builder

Pick a model size, workflow, and budget — get three build tiers with the exact parts to buy.

3 build tiers · savable builds

Plan my build →

Not sure which model?

Model Builder Wizard

Answer a few plain-English questions and get a model, runtime, and stack matched to your goal.

Beginner-friendly · plain-English plan

Get my setup plan →

Want to try it first?

AI Playground

Load a small open model right in your browser and test prompts before you install anything.

WebGPU · nothing to install

Open Playground →

Already running local AI?

Local AI Benchmark

Measure what your machine really does — speed, load time, and whether anything is quietly throttling it.

~5 minutes · nothing leaves your machine

Benchmark my setup →

Want to see inside your setup? New

Command Center

See your real hardware, which installed models fit it, and what Ollama has loaded right now.

Open source · runs 100% locally

See my setup live →

Real output

What the tools actually give you

Real output from our own tools — not mockups. Screenshots are re-captured whenever the checker catalog or scoring changes.

Compatibility Checker verdict card: Qwen3 14B, 14B parameters, 32K context, rated Comfortable with score 91 at Q4_K_M quantization on a 12 GB GPU — 8.9 of 12 GB used with 3.1 GB VRAM to spare — with a copyable 'ollama run qwen3:14b' install command and an LM Studio alternative. — Compatibility Checker verdict · Example · captured July 2026

Model Builder Wizard recommended setup: Qwen3 14B as the general assistant model, Phi-4 as the higher-quality option, Open WebUI or LM Studio as the runtime path, plus deployment-engine and operating-system toggles ahead of a personalized step-by-step roadmap. — Model Wizard plan · Example · captured July 2026

Browser AI Playground with Llama 3.2 1B Instruct loaded and a green MODEL READY badge: the status line reads 'Llama 3.2 1B Instruct is ready. Prompts run locally in this browser session', with a WebGPU adapter detected, Reload and Reset controls, and starter-prompt chips above an empty chat. — Browser Playground, model loaded · Example · captured July 2026

Latest sightings

This Week’s Radar Drops

Fresh finds from the directory, reviewed by our editors.

Browse the full directory →

Updated 2 days ago

Local Deep Research

Run recursive, multi-step agentic research loops completely locally and encrypted using Ollama or llama.cpp.

Read the reviewed profile →Updated 4 days ago

whichllm

Analyze your local hardware topology and instantly see which open-weight models will run comfortably based on real benchmarks.

Read the reviewed profile →Updated: July 2026

llamafile

The absolute friendliest way to run local AI. Turns an entire model checkpoint into a single, zero-dependency executable file.

Read the reviewed profile →

Recently reviewed

GLM-5.2 (753B MoE)Ollama v0.30 VibeThinker-3B (64K Window)Codebase Memory MCP Gemma 4 E4B Mistral Small 4 DeepSeek-V4-Pro Qwen3 14B (Fast)GPT-OSS 20B

Browse by workflow

View all workflows →

Local Chat & AssistantsStart here

RAG & SearchPrivate docs

Coding & IDERepo help

Agents & AutomationMCP workflows

Image Gen & VisionModel workflows

Audio & SpeechWhisper guide

Inference & ServingDeploy models

Data & Vector DBRAG storage

Featured ecosystem tools

Third-party open-source tools our editors have reviewed — the runtimes, interfaces, and databases you’ll run your planned stack with. Not built by us; tested and profiled by us.

View all tools →

Ollama

Run open models locally with a simple CLI, model library, desktop app, and local API.

LocalMIT

Reviewed profile →

Open WebUI

Self-hosted AI platform and chat interface for Ollama and other model backends.

Self-hostedBSD-3-Clause

Reviewed profile →

LM Studio

Desktop app for discovering, downloading, chatting with, and serving local LLMs.

LocalProprietary desktop app

Reviewed profile →

Qdrant

Open-source vector database with strong filtering and production-oriented retrieval features.

Self-hosted or cloudApache 2.0

Reviewed profile →

vLLM

High-throughput open-source LLM serving engine for production and research workloads.

Self-hosted serverApache 2.0

Reviewed profile →

Continue

Open-source AI coding assistant for VS Code and JetBrains with configurable models and context.

Local or cloud modelsApache 2.0

Reviewed profile →

Why trust this

Built on transparency

Source-aware pages, commercial labels, and practical next steps.

Open-source firstCommercial tools labeled

No paywallsFree research paths

Citation-mindedSource notes where useful

Editorial standardsNo hidden agendas.

Read standards →

Foundation: Three concepts that unlock everything else

What is a local LLM?Understand the core architectural and privacy differences between local open-weight model weights and centralized cloud vendor APIs.What is VRAM?Compute your active graphics hardware requirements and allocate a safe baseline layer for active context token KV-cache growth.What is quantization?Learn how bit-depth file weight compression allows large frontier model architectures to run smoothly inside tight consumer hardware memory limits.

All 6 foundation concepts →

Popular comparisons

Ollama vs LM StudioChoose between a scriptable local model runtime and a polished desktop app for testing open models.Local AI vs ChatGPTUnderstand when to run open models locally and when a hosted assistant is still the practical choice.Llama vs Qwen vs Mistral for Local AICompare three major open-weight families for local AI builders.Qdrant vs ChromaCompare a production-oriented vector database with a lightweight RAG prototype favorite.

View all comparisons →

Keep exploring

Go deeper once you have a plan: browse ready-made stack recipes, the knowledge base, and model profiles.

Browse stack recipesMove from tool discovery to implementation.Knowledge baseA 7-topic reference index for local AI.Explore model profilesCompare open-weight models and their trade-offs.

Common questions

Honest answers before you start

Do I need a GPU to run local AI?

No — small models (1–4B parameters) run on a normal CPU, just slower. A GPU with 8 GB+ of VRAM is where mid-size models get comfortable. The Compatibility Checker tells you exactly where your machine lands.

Does my data leave my browser?

The specs you enter in the Checker and PC Builder stay in your browser — we never upload them. The Playground runs its model locally in your browser via WebGPU. Analytics only runs if you accept it in the consent banner.

Is this really free?

Yes. All six tools are free with no account. The site earns money from clearly labeled affiliate links on some commercial tools — never from your usage of the tools or paywalled content.

Can my laptop run a local model?

Probably — most laptops from the last few years can run a small quantized model. Whether it feels good depends on RAM and (if you have one) GPU memory. Run the checker above; it is honest about “tight” fits.

Do I need to know how to code?

No. Tools like Ollama and LM Studio are install-and-run, and our recommendations come with copy-paste commands. If you can install an app, you can run a local model.

What if my hardware is too weak?

You still have paths: run a smaller model, use the PC Builder to plan an affordable upgrade, or rent a cloud GPU for occasional heavy jobs (we label those partner links clearly).

See what your PC can run

Free, no signup, about 60 seconds. Your specs never leave your browser.

See what your PC can run →

Check your PC. Choose your models. Build your local AI stack.

How to use OpenSourcesAI

Local LLM Compatibility Checker

Auto-detect my device basics

Local LLM Hardware Compatibility Matrix

Start with the right tool

Compatibility Checker

Local AI PC Builder

Model Builder Wizard

AI Playground

Local AI Benchmark

Command Center

What the tools actually give you

This Week’s Radar Drops

Local Deep Research

whichllm

llamafile

Browse by workflow

Featured ecosystem tools

Ollama

Open WebUI

LM Studio

Qdrant

vLLM

Continue

Built on transparency

Foundation: Three concepts that unlock everything else

Popular comparisons

Keep exploring

Stay ahead in open-source AI

Honest answers before you start

See what your PC can run