You Don't Need Billions to Build Frontier AI | Marjorie Janiewicz
MTS
October 30, 2026
Mistral CRO Marjorie Janiewicz joins MTS to break down the open-weight release of Mistral Large 4.
Marjorie Janiewicz
You don't need billions of dollars to build a frontier model that's as strong as closed source models. And that value that we unlock for our own model training, we bring that back to our customers. Customers are starting to worry about their bills, the token price and how much investment is going into their AI.
When Mistral is able to really build those models at a fraction of the cost, it does mean that it provides customers the ability to also leverage smaller models that are more efficient. And in turns, they are going to see some of that value back in the cost efficiency by leveraging Mistral.
Sophia
Hello everyone, welcome back to MTS. Today, we are joined with Marjorie Janiewicz, who is the Chief Revenue Officer at Mistral. And today, they just made their weights frontier open model Mistral Large 4, ML4 for short, accessible, widely accessible. So Marjorie, welcome to MTS.
Marjorie Janiewicz
Thank you very much. Thank you for having me. Exciting day.
Sophia
Big exciting day. So for people who are just following along, what does this announcement mean?
Marjorie Janiewicz
It's a big moment, Sophia. So Mistral Large 4 is our latest flagship model. It's the best open source model in coding. It's state-of-the-art in cyber security. It's best of breed in financial services and manufacturing. And it's open weight. It means that it's available to everybody. It's a big moment for companies and governments alike trying to achieve sovereignty and autonomy in AI.
So companies no longer have to choose between closed source and frontier.
Mistral and a fraction of the GPU investments just launched a really incredible model. So we're very happy. Yeah.
Sophia
You guys specifically share that your state-of-the-art in cyber defense capabilities. Tell me a little bit more about that. What does that mean?
Marjorie Janiewicz
Yeah. So I'm sure your listener have noticed that over the past few months there's been quite a bit going on in cyber attacks. Quite a bit indeed.
Basically, either models, closed-source models disconnecting really enterprises from access to their models, but also enterprises starting to suffer from attacks from closed-source models. So really our customer base has been asking us to really get access to open-weight models that are very good at cyber so they can really protect themselves against those cyber defenses, those cyber attacks. So that's what the model is very good at. So in the announcement today, we released really this model open-weight. It's very good at cyber. It means that whether you are a financial services institution wanting to really run a cyber model, really on their own premises to be independent or a cyber vendor who want to really explore how they can leverage that model to really help the world becoming safer, really Mistral Large 4 is a great choice for that. So we are very, very excited.
Sophia
Yeah, I mean, we shared earlier, I mean, during the Hugging Face incident, famously Hugging Face actually had to use an open-weight model to defend against these closed models that were attacking it, and they didn't know exactly what was happening, why. I mean, you guys also shared, this is one of the best open models built in the West. It ranks in the top 10 top open models in the world for agent decoding on the Artificial Analysis Coding Agent Index.
What other things can you share about why this is one of the best open models built in the West?
Marjorie Janiewicz
Yeah, so really, the model is the best open model in coding period and it's open.
It means that for companies that are relying nowadays on coding, they don't have to choose between basically a closed-source model and another model in China that's good at coding.
Right now, Mistral AI Large 4 is the best open model across coding overall. It can be accessible to all and it's way more efficient. So it's less token consuming versus other models as well. So it does mean that companies have now an alternative. They don't need to rent the model from a third party. They can own it. That's really the promise of open weight models. It's really being able to own those models. And when that model is best in class at coding and it's open, I think it's really providing a net new set of opportunities for companies that are trying to achieve autonomy and do not want to choose between frontier capabilities in coding, and basically renting those from expensive closed source vendors.
Sophia
Yeah. I want to talk about that even more because especially when it comes to enterprises outside of just the United States, I want to talk about Globally all across the West, you're seeing this common theme where enterprises want sovereignty. They want to be able to use frontier intelligence, but without being locked into token prices or models or have their entire businesses dependent on it. So tell me a little bit more about that from what you're hearing on the ground as you guys work with these enterprises. Yeah.
Read the other 15 minutes of this episode here.
$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.
Not what you expected? Full refund within 14 days.
Thousands of transcripts fetched by people building searchable podcast archives
Fetch transcripts with code
The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000798897089