Mark Hadley

Jan - Open-source ChatGPT alternative that runs 100% offline

by

Jan runs 100% offline on your computer, utilizes open-source AI models, prioritizes privacy, and is highly customizable.You can download Jan for Windows, Mac or Linux. Mobile coming soon.

Add a comment

Replies

Best
Mark Hadley
Hunter
📌
Im just a user. User-Friendly Interface: Jan.ai’s design makes interaction a breeze. No steep learning curves—just dive right in! Cross-Platform Compatibility: Whether you’re on Windows, Mac, or Linux, Jan.ai has you covered. Multiple Models: Access a repository of LLM models effortlessly. The power of choice at your fingertips. ChatGPT Support: Jan.ai’s integration with OpenAI’s API brings ChatGPT into the mix. 100% Free & Open Source: Jan.ai is committed to democratizing AI. No paywalls, just pure accessibility. Offline Use: Take Jan.ai wherever you go—even when the internet doesn’t. Offline support for the win! 🌐 Kudos to the Jan.ai team for creating a rad project! 🤖🎉 hardware requirements: https://jan.ai/guides/install/ha... discord: https://discord.gg/FTk2MvZwJH they are hiring: https://janai.bamboohr.com/careers
André J
Love the simplicity! Having followed this space closely for the last year. Two questions comes to mind: 1. What specs do you need to run this? 2. And how does it compare to GPT 3.5 and GPT 4?
Mark Hadley
@sentry_co 1. Hardware requirements can be found here https://jan.ai/guides/install/ha... 2. It depends on which LLM model you download. There are many so not an easy comparison. You can also use both ChatGPT using OpenAI's API within the app so if you prefer those LLM but that requires you to be online.
André J
@mrkhdly Thanks. No need to compare directly. But Have you tried benchmarking them against ChatGPT 3.5 and 4? Like ther must be some standard benchmarking out there right?
Mark Hadley
@sentry_co I'm just a casual user / hobbyist and haven't done any sort of structured benchmarking. If you'd want benchmarking results looking up the models on hugging face or searching on Reddit might be the best. Sorry I wasn't of help to your question.
André J
@mrkhdly I think it would be possible to do some low-tech benchmarking against the competition. Like ask the top most 10 common questions people ask GPTs. And compare the result. Users decide to try a product if they can determine if a product is better than another product they are already using. That's why I am asking. If this product is even close to the competition then adoption would sky rocket. But right now it's hard to tell from the marketing material. But I don't think it's hard to produce such data.
Mark Hadley
@sentry_co If you’d like you can produce this data & contribute it to the models.json files https://github.com/janhq/jan/tre... project is open source
Nicole Zhu
Hi, member of the core team here. Thank you for hunting us! What a lovely surprise to wake up to on Sunday :) Linking our Discord here for folks who need troubleshooting or feature requests: https://discord.gg/DmrfNEUZCY
Saarth Shah
Looks promising, is this going to be completely open source? Just curious what the business model looks like?
Mark Hadley
@saarth_shah1 Source repo is here: https://github.com/janhq/jan & I don't know what their future plans are for a business model but currently it is bootstrapped https://jan.ai/about/#bootstrapped
Garen Orchyan
Looks pretty cool. Congratulations on the launch and good luck♥️🦄
Peter Puhov
It looks very promising! I reviewed the Git stats and found that this project has over 30 contributors and is highly active. I was very impressed and will test this repository today.
Sankar Isanaka
Interesting, is there a doc/URL to read more about the tech? Congratulations team!
Mark Hadley
@sanhq Here is a link to their documentation on the framework for the app https://jan.ai/docs/
Sahukari Bhargavi
amazing idea congratulations on your launch
Larry Hartman
Congratulations on an amazing offline solution.
Nuno Reis
Oh nice! Is it just text for now? And what are the minimum specs to run it?
Ivan Somov
Shout out to the philosophy of the author and app! Congrats on the launch!
Dmitrii
Jan is a software product that appears simple, but its capabilities are limitless. Key Features: 1. Open-source product: This product's open-source code ensures transparency in the development process, as the Jan.ai team works cohesively and passionately, allowing for continuous expansion of its capabilities. 2. Architecture: Jan consists of two components - the UI (Jan with dialogues) and Nitro (a extension for loading local models). This approach is advantageous as it allows more flexibility in tackling tasks. For instance, if I need something that's not available in Jan, as a developer, I can directly use Nitro, keeping me within the same stack rather than switching between several. 3. API: Jan supports an API that is compatible with OpenAI. This is a competitive advantage for those who wish to transition from cloud-based models like ChatGPT to local ones as the adaptation process would not require significant resource investments. Take note that Jan is designed for the usage of AI models in offline mode, which means certain tasks previously resolved by cloud-based AI owners would need to be done. 1. Finding suitable hardware. In reality, a modern laptop from Apple or a PC with a fast CPU and (or) fast graphics card will suffice. A lot depends on the AI models - you can run an extremely "simple" model on a mobile processor or something more advanced, comparable to ChatGPT 3.5, on a group of powerful graphics cards. The development of Jan at the time of my review is actively ongoing, so it's currently impossible to list all compatible hardware. I am certain that support will eventually cover a wider range of devices. 2. Choosing an AI model. This is quite debatable because the number of available models is significant, and new ones appear regularly. Some resemble ChatGPT 3.5 in tests while others are good at solving only specific tasks. This isn't something Jan is obligated to help with. What's important is that it allows the user to launch arbitrary AI models, and this is its true value. As a user, I prefer to use Jan specifically because it allows me to be independent from cloud AI models. I can be assured that the confidentiality of processed information will not be compromised. It's gratifying to realize that what you send in the dialog box will not be used for analyzing model behavior, will not be used to train other AI models, and will not be used for marketing purposes.
max
Awesome product! Works flawlessly, simplest local AI I've installed so far. Missing some key functionalities though like image upload, document etc.
Arun Kodavali
Have been using this for some time, looks great!
Ivy Compeh
Congrats to the maker for creating such a genius, open-source product. It is truly a game-changer, am I right? Keep killing the game!
Abdul-Mujib Aliu
This is a tidy solution.
Egor Karpov
Looks quite promising. Best of luck to you today 🙌
Nisa
Congratulations for the launch... Amazing product
Dilshan Samudu
Wow this is the great project keep it up.