4 min read

DeepSeek R1 is Free on UIUC.chat - Product Update #3

You’re receiving this email since you’ve signed up for UIUC.chat. We will only email you about product updates. This is our 3rd email in our two-year existence, but we have 1,700 git commits and 600 PRs across our frontend and backend. I hope this is ‘high alpha’ insight into the best uses of LLMs that we’re pushing into production. We are a small team of researchers and everything we build is free, enjoy!

👉
UIUC.chat is the easiest way I can imagine to train your own LLM, then share it like a Google Doc.

Today we're launching:

  1. DeepSeek R1 is FREE on UIUC.chat!
  2. Qwen VL 72B, the best Vision Language Model, is FREE on UIUC.chat!
  3. New Dashboard page with Sharing like Google Docs
  4. New prompt engineering controls and features

And last month we silently launched Guided Learning (Aka Tutor Mode), a whole-new Analysis page and Claude support. That post is well worth a read.

DeepSeek R1 is Free

UIUC.chat now offers free access to DeepSeek R1, a powerful open-weight LLM optimized for reasoning and research. Whether you're tackling complex coursework or need an AI assistant for your studies, you can now leverage DeepSeek R1 at no cost.

To use it, simply go to any project on UIUC.chat, or make a new one, and select that model in the chat window.

We're expanding our self-hosted offerings. We keep updating this as new models are released.

Free VLMs hosted at NCSA

We're now hosting Qwen VL 72B, the best Vision-Language model today. It's extremely powerful, and now available for free to all users.

This uses 8x A100 (40GB) GPUs to serve its large context window.

We self-host these models here at Illinois NCSA, and in the spirit of being a public university, I'm proud to share them publicly to give back to our community and to remove any barrier from people enjoying our app.

Qwen 2.5 VL just launched... waiting on software support

🧠
A side note: Qwen just released the latest versino of their Vision models, Qwen 2.5, we'll have it online in the next few days (we're waiting for vLLM support).

Here's a few of our favorite tweets on Qwen 2.5 VL.

Hopefully Qwen strikes a balance between great benchmark results and extremely plesent practical usage.

We're excited to build computer-use features into UIUC.chat.

New Dashboard page

Our new Dashboard has two key goals:

  1. Make it easier to use our connectors, like Canvas, GitHub, MIT, and website crawling.
  2. Elevate the "Share Chatbot" controls, where you can add users and admins to chat with or edit your project.

I'm extremely proud of our delightful Google Docs-like sharing controls.

  • Decide if you want your project publicly viewable or totally private. Either way, you can add edit-capable admins to collaborate on building your chatbot.
  • If it's private, you may add view-only users who will only see your Chat page, but none of the dashboards.
You can paste messy lists of emails (like those coppied from Outlook) and we'll capture all the addresses no problem.

We're delighted by the redesign to make document ingest even more intuitive.

The new Dashboard page. The "Share Chatbot" button is how you add users and admins to your project.

New Prompt Engineering Features

Now on the Prompting page, you can share customized links that will enable certain prompting features just for that link. This makes it easy to have multiple versions of your chatbot for different users and purposes.

For example, you can share a version of the chatbot that has Guided Learning enabled, which we launched last month. Or you could share a version that focuses the chatbot on your documents only, thereby limiting its outside world knowledge, just for specific critical tasks like reviewing a contract or other legal documents.

The goal is to make your chatbot as customizable as possible, supporting many different users.

Previously, these same controls had to be applied globally to your chatbot. Now they can be applied on a case-by-case basis, giving more flexibility and control to project owners.

Sharing customized links that alter the behavior of your chatbot. Useful for users to have specific control over how their bot behaves on a case-by-case basis.

Thank you, and if you have any feedback or feature suggestions please just reply to this email and I'll build it for you.

And if you made it this far in the email, consider checking out our Patreon to support us starving students. Everything we build is free thanks to the University of Illinois, NCSA and the NSF.