grock has just added their first reasoning model to their platform deep seek R1 distill llama 70b is now available I'm just going to quickly show you a demo of what I plugged into an application that I built a little while ago where you can see for yourself it will run through the process generate the reasoning tokens and you can be the judge for yourself just how fast this model is here is just a really quick demonstration of the model in action as you see it just flies through that Chain of Thought and in addition to
that you can see that this is a very coherent response so if I ask a follow-up question with several different changes you'll see it will apply that as well to try this out you can go to console. ro.com you can select deep seek from the model dropdown and you can ask it whatever you like if I just send in the infamous how many RS are in Strawberry question what you'll see back is the thinking within this XML tag and then you'll see the response at the bottom here the word strawberry contains three AR and if
we go through the thinking process you'll see all of that Chain of Thought here which is actually quite entertaining to read if you haven't tried out the Deep seek chat interface as well the one thing that I do want to touch on because this will probably be a question that some people have when deep seek R1 was released they released a mixture of experts model that has 671 billion parameters or 37 billion activated parameter when they had this release they also released a handful of other models ranging ing from 1.5 billion parameters all the way
through to 70 billion parameters within those R1 distilled models there were these quen variations as well as these llama variations the model that they're serving up on Gro is this deep seek R1 distill llama 70b of those models that were released this is both the largest as well as as you might assume the most powerful model that's just one thing that I want to point out is this is not to be confused with deep seek R1 which is their larger mixture of xert model this model is the dense deep seek R1 distill LL 70b model
now the one thing that I didn't quite see yet is the pricing when I went to check that out I saw that it isn't quite available yet but in terms of the rate limits so I'm on their paid tier and right now at time of recording it's showing request per minute at 100 50,000 requests per day as well as 30,000 tokens per minute these numbers will VAR account to account but otherwise kudos to the team at grock for the great work as always let me know what you're going to be building with this are you
excited to play around with this what sort of applications are you potentially going to be leveraging this for otherwise if you found this video useful please like comment share and subscribe otherwise until the next one
