As an update to the issue which I posted recently I think the underlying problem was with the way NA dealt with the original query which was done by inserting an image in .png form. There must have been an extraction issue. When the Latex code (or raw input style ) is inserted after about 5 minutes or so the correct answer was output yet when I congratulated it on getting the correct answer it went into seemingly endless circle. So NA can get the right answer but it is also glacial.

It is likely that OpenAI's PDF import is using a combination of reading in PlainText as well as doing processing of the page images using its multi modal capabilities. It is also likely that is bringing a number of other programs to bear on the problem. When I use the gpt-4o model on a image of a PDF with equations it does a quite good job (with the correct prompting) to get the content of the page and include the equations with embedded LaTeX strings. As a general rule in work like this, which I've done a bunch of over the past year, if the image of the page has many equations that are themselves quite small, imagine a grid of equations were the equations have very small font sizes, the LLM will often make a number of errors on the content that has a small font size. Also, if there's a fair amount of very complex, two dimensional layout of equations, or different equations laid out next to one another that should be distinguishable from one another, then it will often times make such errors. I have not tested this on open AI's website, all of my work has been within Wolfram language itself. I suppose it's also worth remembering that the import of a PDF using the Import function has all of the restrictions of any other importing such as the various python packages, as far as I understand it. And in Wolfram language, the import function will import plane text and often times, if the text is laid out in a way that is not strictly linear be confused about which pieces of text relate to other pieces of text for example different pieces of text in a grid. Properly prompted vision models will be able, sometimes to make the distinction. But it all depends upon very sophisticated prompting. And for a sufficiently complicated layout of a PDFs page, as well as the level of detail of equations, and so on, will lead to errors.

Groq is supported.

Actually, It's pretty well known here, but very easy to mistake. ;-) There is no ServiceConnect yet for XAI in the Mathematica distribution. It's worth making a suggestion that it be added and see whether the developers have it in their plans to add it or, if not, what the reason might be.

Will do!

OK, found the place to get connected. Turns out it is all automatic. Magic!!

I watched Stephen's introduction to the Notebook Assistant and I will have a play with it. The way it is marketed suggests I don't have to buy tokens from OpenAI for the lowest level of the product. It is USD 240 and that covers what the LLM is doing in the background. I assume you have done a behind the scenes deal based on assumed levels of non-industrial strength usage and if you want to really get serious you buy Wolfram's service credits. Have I got it right?

Thanks Bradley. I have been a user of Mathematica since probably 2.0 and I was discussing the potential use of the NA with my oncologist daughter who will be doing some bioinformatics at Cambridge and Stockholm in 2025 as part of her PhD and the NA is just the sort of tool that could help her. By the way, I loved Stephen's warts and all presentation. I'll get playing with Notebook Assistant. Cheers Peter

Peter I have the same impression. Take the more trivial case of Canvas which replaced the awful Drawing Tools. It's a step forward, yes, but needs some effort by senior developers in order to improve the design and sort out the bugs. I called Tech Support in April (2025) to register my frustration with canvas and was told my name would be added to the list of complainers. The funny thing was after the call I realized I was editing a sketch that I'd created in around 2020. I had similar problems then and am pretty sure I called to make the same compliant. Hopeless!

FYI - ten years ago Mike Horneychurch got frustrated with Wolfram and suggested they spend a year just sorting out the backlog of bugs instead of flitting from one flower to the next.

Hello Héctor, thanks for the question.

Yes, I can confirm that all of the Notebook Assistant and LLM Kit features and capabilities are included in both the Essentials and Pro subscriptions. The difference between the two subscriptions is only in the usage limits. The Pro subscription has double the usage limit of the Essentials subscription.

Hope that helps!

Great to hear from you, @John ! If you stumble on some neat examples - please share :-)

There was a bug which I reported to the developers and I believe they have pushed out an update that should fix this. If you continue to have this issue, you should contact technical support.

You are welcome. It was something close to my heart because internally I was the first person to run across this :-). And the developers were very responsive in getting it fixed as quickly as possible.

Impossible not to enjoy the dinosaur story from Stephen :-) A quote from https://wolfr.am/NotebookAssistant

And that reminds me of the very first time I ever saw a computer in real life. It was 1969 and I was 9 years old (and the computer was an IBM mainframe). The person who was showing me the computer asked me: “So what do you want to compute?” I really had no idea at that time “what one might compute”. Rather lamely I said “the weight of a dinosaur”. So, 55 years later, let’s try that again: