AI Compare Ninja AI Compare Ninja

Compare AI tools. Use prompts and skills.
Language
EN

Local AI Tools Compared

Local AI runs on your own computer instead of in a provider's cloud. You download an open model and use it with an app or program, often without internet. Here you can compare suitable apps and models, with sources.

Filter by subcategory

All 14Apps & interfaces 5Runtimes & servers 4Open models 5

Sort

AnythingLLM

Apps & interfaces

Desktop app for chatting with your own documents and AI agents that runs locally on your computer; no account needed.

Free accessFree plan
Paid fromnot published
  • For macOS (Apple Silicon and Intel), Windows (x64 and ARM) and Linux; open source under MIT
  • Local by default per the vendor: models, documents and chats stay on the device

Details Visit site (opens in a new tab)

ComfyUI

Apps & interfaces

Free open-source software for AI images and videos on your own computer; you build workflows from connected steps.

Free accessFree plan
Paid fromnot published
  • Desktop version free forever and open source under GPL-3.0, per the vendor
  • Fully usable offline after setup, per the vendor

Details Visit site (opens in a new tab)

Foundry Local

Runtimes & servers

Free runtime from Microsoft that lets developers build AI models into their own apps that run directly on the device.

Free accessFree plan
Paid fromnot published
  • For Windows, macOS (Apple silicon) and Linux; per the vendor no Azure subscription and no per-token costs
  • Automatically picks the device's NPU, graphics card or CPU, per the vendor

Details Visit site (opens in a new tab)

Gemma

Open models

Open model family from Google DeepMind that you can download and run on your own device; Gemma 4 is released under Apache 2.0.

Free accessFree plan
Paid fromnot published
  • Per the model card, Gemma 4 comes in five sizes: E2B, E4B, 12B, 26B A4B and 31B
  • Per Google, E2B and E4B are for phones and laptops, 12B and up for graphics cards and workstations

Details Visit site (opens in a new tab)

gpt-oss

Open models

Open language models from OpenAI to download, made for desktops, laptops and data centres; under Apache 2.0.

Free accessFree plan
Paid fromnot published
  • Sizes 20B and 120B, plus gpt-oss-safeguard for your own safety rules
  • Per the model card, gpt-oss-20b runs within 16 GB of memory, gpt-oss-120b on one 80 GB graphics card

Details Visit site (opens in a new tab)

Jan

Apps & interfaces

Free desktop app for using open AI models directly on your computer, even fully offline.

Free accessFree plan
Paid fromnot published
  • For Windows, macOS and Linux; open source under Apache 2.0 (per the licence file)
  • Per the privacy page, chat history and usage logs stay on the device

Details Visit site (opens in a new tab)

Llama

Open models

Open model family from Meta to download; Llama 4 is released under Meta's own community licence with conditions.

Free accessFree plan
Paid fromnot published
  • Per Meta, Llama 4 is natively multimodal (text and image) with a mixture-of-experts design
  • Per the licence: if you pass on a product with it, include the licence and display “Built with Llama”

Details Visit site (opens in a new tab)

llama.cpp

Runtimes & servers

Open-source software that runs language models on your own computer, as a small app for Mac and Windows or from the command line.

Free accessFree plan
Paid fromnot published
  • Free, open source and usable offline per the vendor; app for Mac and Windows 11, command line for Linux
  • Per the project, a CPU is enough; graphics cards are used via Metal, CUDA and Vulkan, among others

Details Visit site (opens in a new tab)

LM Studio

Apps & interfaces

Desktop app for running open-source LLMs locally.

Free accessFree plan
Paid fromCloud credits (optional) · pay as you goCloud inference for frontier models
  • Run open-source LLMs locally on the desktop
  • Bionic Agent for local agent workflows

Details Visit site (opens in a new tab)

Mistral (open models)

Open models

Open models from Paris-based Mistral AI to download, such as Ministral 3 and Mistral Small 4, many of them under Apache 2.0.

Free accessFree plan
Paid fromnot published
  • Per Mistral, under Apache 2.0: Mistral Small 4 and Ministral 3 in 14B, 8B and 3B
  • Per the model card, Ministral 3 14B fits in 24 GB of graphics memory (FP8)

Details Visit site (opens in a new tab)

Ollama

Runtimes & servers

Open-source tool for running language models locally, complemented by a paid cloud service for developers and teams.

Free accessFree plan
Paid fromCloud Pro · 20 $Access to larger Pro cloud models and 3 concurrent requests
  • Open-source tool to run language models locally such as Llama, Gemma, Mistral, and Phi
  • REST API compatible with the OpenAI API for easy integration with existing apps

Details Visit site (opens in a new tab)

Open WebUI

Apps & interfaces

Self-hostable AI interface for local and external models, knowledge sources, tools, and team access.

Free accessFree plan
Paid fromEnterprise licence · contact salesWhite-labelling and customised branding
  • Self-hosted chat UI for local and external models
  • Knowledge sources, tools, and team access in one install

Details Visit site (opens in a new tab)

Qwen (open models)

Open models

Open models from Alibaba's Qwen team to download, in many sizes; licence per model is Apache 2.0 or the Qwen licence.

Free accessFree plan
Paid fromnot published
  • Example Qwen3.8-27B: under Apache 2.0 per the model page
  • Other models are released under the Qwen Community License with its own conditions

Details Visit site (opens in a new tab)

whisper.cpp

Runtimes & servers

Open-source software that turns speech into text locally, using OpenAI's Whisper speech-recognition models.

Free accessFree plan
Paid fromnot published
  • Open source under MIT; for macOS, Linux, Windows and other platforms
  • Usable on a CPU alone per the project; graphics cards from NVIDIA and AMD, among others, are supported

Details Visit site (opens in a new tab)

Local AI explained

When AI on your own computer makes sense, hardware, privacy and how to start in 3 steps.

Local AI: overview

Prompts, skills and agents for this topic

Templates for this topic. Each one links at least one specialist tool from this category.

Terms explained

Local AIOpen-weight modelQuantizationGPU (graphics processing unit)

Alphabetical order. This is not a quality ranking.

Comparison criteria

We compare free access, paid entry, plan limits, features and the listed official sources. Unsourced figures stay marked as not published.

Methodology

Only confirmed entries appear in the catalog. Prices, plans and limits come from listed official vendor pages. Missing figures stay marked as not published. Search results are ordered by Task match by default; category and task lists are alphabetical. Both can be switched to Free access first; neither is a quality ranking. The vendor page always wins.

Is this a ranking?

No. Search results are ordered by Task match by default; category and task lists are alphabetical. Both can be switched to Free access first. We do not award stars or a test winner.

What does "free access" mean for the listed tools?

Free access can mean a permanently free plan, a time-limited trial, or a one-off starting credit with no card required. The exact type is stated on each tool's own page.

How many of the listed tools offer free access?

14 of 14 listed tools offer free access - a permanent plan, a trial, or a no-card credit.

How do I see tools with free access first?

Use the “Free access first” toggle above the list to see tools with free access first.