review galore

Review · Local AI

Jan review

A free, open-source desktop AI app can suit a budget-conscious team—but it’s not a shared team workspace.

6.8/10Review Galore score

We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money

The bottom line

A $0 way to run local AI, if your team is comfortable setting up models on its own machines.

Small teams that want to run open AI models on desktop computers can try Jan without a subscription. Its free plan lists unlimited local inference and model downloads; connecting online models requires your own API keys, with provider charges billed separately.

Jan may be a poor fit if you need shared team controls or a polished, low-setup workflow. Its homepage says memory features are coming soon, and a Reddit discussion includes both enthusiastic local-use reports and an account of an external-drive setup breaking.

Good fit if

  • Budget-conscious local AI users
  • Small teams choosing their own models
  • People comfortable managing API keys

Skip it if

  • Teams needing shared AI workspaces
  • People avoiding desktop setup
  • Teams relying on built-in memory today
Jan homepage
Jan homepage, captured Nov 11, 2026

ScoresHow it rates

  • Features7
  • Ease of use5
  • Value for money9
  • Support5
  • User sentiment5

What works

  • Free plan lists unlimited local inference
  • Unlimited model downloads are included
  • Apache-2.0 open-source license
  • Choose open models or online providers
  • Online connections use your own API keys

What doesn't

  • Online providers bill you separately
  • Memory features are still coming soon
  • Hardware requirements are not published
  • Team administration details are not published
  • A Reddit user said an external-drive setup broke

01Who Jan is for

Jan makes sense for a one-to-ten-person team that wants to try AI without adding a software subscription. It’s a desktop app for chatting with open models locally or connecting to online models. The homepage lists providers including OpenAI, Anthropic, Google, Meta, Mistral, Alibaba, DeepSeek, and Moonshot AI.

The local option may suit a small team handling experiments or everyday prompts on its own computers, especially if people want to choose their models. Jan’s homepage describes the app as free and open source, and its pricing lists unlimited local inference and model downloads. Online connections need your own API keys, and providers bill you directly.

This is not an obvious pick for a team that needs shared conversations, central administration, or consistent access across staff. Jan’s homepage focuses on a desktop app, and it says memory features are coming soon. The homepage lists a Mac download; check platform availability before planning a rollout on other systems.

02Jan free plan and trial

Jan’s free / open-source plan costs $0 per month. The pricing page lists unlimited local inference, unlimited model downloads, and an Apache-2.0 open-source license. There is no paid tier listed, so there is no higher plan to unlock more local usage.

A free trial is not listed because the app itself is free. The catch is online usage: connecting to a provider requires your own API keys, and those third-party providers bill you directly. So Jan’s $0 price does not cover online model usage. Jan’s homepage also says memory features are coming soon, so don’t count on that planned feature being available now. For a team that wants local models and accepts desktop setup, the free plan is the cheapest way in.

03What a small team pays for Jan

The listed price for Jan’s free / open-source plan is $0 per month. That makes the monthly Jan software cost $0 for one person, $0 for three people, or $0 for five people. This is a free app, not a per-seat subscription, so the price does not rise with headcount. Multiplying the monthly listed price by 12 gives $0 for a year at any of those team sizes, though Jan does not publish an annual billing price.

The plan lists unlimited local inference and unlimited model downloads, with an Apache-2.0 license. Jan has no paid tiers listed. The extra cost to watch is online model use: Jan says connections use your own API keys, and third-party providers bill you directly. Jan does not publish those providers’ rates as part of its own plan price, so ask each provider what your team’s usage would cost before connecting it. There is no published Jan add-on price to add to the $0 app cost.

04Getting started with Jan

The basic path is to download Jan Desktop, choose an open model to run locally, or connect an online provider using your own API key. Jan’s homepage lists a Mac download and names supported provider options, but it does not publish hardware requirements there. Check that your computers can handle the models you plan to use before making Jan part of a daily workflow.

For local use, model downloads are included in the free plan. A Reddit user described running Jan offline, while commenters noted that local model setup and hardware matter. Another user said an external-drive setup later broke, so don’t assume a portable setup will be trouble-free. For online use, your team must manage its API keys and separate provider billing. Jan says it is built in public, but setup support details are not published on its homepage. A small pilot on one computer is a sensible first step.

05When small teams outgrow Jan

Jan’s free price is hard to outgrow, but a team can outgrow the workflow. Its homepage describes a desktop app and does not publish team administration or shared-workspace details. If you need centrally managed access or shared conversations, ask Jan whether those capabilities are available before standardizing on it.

Online usage can also become less predictable than the $0 app price suggests: each provider bills you directly for API use. Jan’s homepage says memory features are coming soon, so teams that need context carried across conversations may need to wait or choose another workflow. Reddit reports are mixed: one poster praised offline use, while another said an external-drive setup broke. Those are individual experiences, not a guarantee about every setup. Teams should test their intended model and computers before depending on Jan for routine work.

Questions people ask

Is Jan free for a small team?

Yes. Jan’s pricing lists a free / open-source plan at $0 per month, with unlimited local inference and unlimited model downloads. The app price stays $0 whether one, three, or five people use it. Online model providers bill separately when you connect with your own API keys.

Does Jan work without an internet connection?

Jan supports local inference, and a Reddit user reports running it offline. That does not mean every part of setup is offline: you may need to download models first. Check your intended model and computer setup before relying on Jan without internet access.

Which AI providers can I connect to in Jan?

Jan’s homepage lists OpenAI, Anthropic, Google, Meta, Mistral, Alibaba, DeepSeek, and Moonshot AI. Online connections use your own API keys, and those providers bill you directly. Check provider pricing separately before estimating the total cost for your team.

Does Jan have memory features?

Jan’s homepage says memory features are coming soon. That means you should not plan around memory carrying over context and preferences as an available feature today. If persistent context is essential to your workflow, ask Jan for its current availability and timing.

Is Jan available for Mac?

Yes. Jan’s homepage lists Jan Desktop as available to download for Mac. The published homepage details do not establish availability for other operating systems, so check Jan’s current download page if your team uses Windows or Linux.

Sources

  1. Jan homepage
  2. Jan pricing, compiled via web search from kunavo.com, hokai.io, promptquorum.com, saaskart.co, agentsai.fyi, toptoolspick.com, weavai.app
  3. Reddit: Is Jan AI a virus ?
  4. Reddit: I MANAGED TO GET JAN AI ON AN EXTERNAL HARD DRIVE, OFFLINE RUNNING LLAMA.CPP!