Jan

Orrin Batz

Congratulations on your launch!

10 months ago

Godfrey Donnelly

Nice product so far. Installed cleanly. Having a spot of bother with the tiny llama model (something is amiss), but now downloading Mistral to see if that's better. GPT4 works fine out of the box once the API key is installed. It looks very promising, and the fact that it's open source and multi-platform is a real bonus. I'm assuming that the mobile version will use a local server? If so, that could be a real game changer!

11 months ago

Rocio Cormier

great project.. keep success always..

1 year ago

Franz Bauch

Easy to set up and use on Ubuntu. 16GB of RAM is enough to run Mistral 7B. Generates at a speed of 200 words per minute. Obviously not as fast as using a cloud provider but I can see some uses for it, especially for asynchronous processes.

1 year ago

Paul Kshlerin

Cool! Eager to test it in website- search

1 year ago

Buck Yundt

Congratulation's for launch , Looks promising

1 year ago

Quincy Medhurst

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: <a href="https://discord.gg/DmrfNEUZCY" target="_blank" rel="nofollow noopener noreferrer">https://discord.gg/DmrfNEUZCY</a>

1 year ago

Felipe Sanford

Amazing 🤩

1 year ago

Lamont Strosin

This sounds great! I appreciate the focus on privacy and the use of open-source AI models. The fact that Jan runs offline and is highly customizable is also a big plus.Congratulations on the launch!

1 year ago

Curt Nienow

Comment Deleted

1 year ago

Godfrey Donnelly

Awesome product! Works flawlessly, simplest local AI I've installed so far. Missing some key functionalities though like image upload, document etc.

1 year ago

Roberto Smith

Congratulations on the launch! I really liked your project

1 year ago

Jackson Fisher

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.

1 year ago

Michel Heathcote

Congratulations for your new lunch

1 year ago

Monserrate VonRueden

Nice

1 year ago