# Open-model gaps are shrinking as AI shifts toward specialists

Markdown copy of https://realalpha.app/s/open-model-gaps-are-shrinking-as-ai-shifts-toward/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

- Speaker: [Arthur Mensch](https://realalpha.app/builders/arthur-mensch/) · Arthur Mensch is CEO and co-founder of Mistral. · Co-founded and leads Paris-based Mistral AI as CEO; named to Time's 100 Most Influential People in AI.
- Host: [Alex Kantrowitz](https://realalpha.app/people/alex-kantrowitz/) · Founder, Big Technology · Former BuzzFeed News senior tech reporter and author of Always Day One.
- Show: Big Technology Podcast
- Episode: [Who Wins if AI Models Commoditize? — With Mistral CEO Arthur Mensch](https://realalpha.app/v/xxUTdyEDpbU/), Jan 16, 2026
- Moment: 25:39 to 32:23 of the episode (7 min), playing on https://realalpha.app/s/open-model-gaps-are-shrinking-as-ai-shifts-toward/; the episode from there: https://realalpha.app/v/xxUTdyEDpbU/?t=1539

## Quote

> “If you look at the trend in 2025, I think the gap is more around **three months**.”
> — [Arthur Mensch](https://realalpha.app/builders/arthur-mensch/) at [25:45](https://realalpha.app/v/xxUTdyEDpbU/?t=1545) · Big Technology Podcast

## Takeaways

1. Open models reportedly went from about a six-month lag to three months.
2. Broad reasoning may improve without creating a durable lead.
3. Specialist models can cost less to deploy at scale.

## About this segment

Arthur Mensch says the open-versus-closed performance gap narrowed from about six months to three as compute spread. He expects future differentiation from experts training models for specific fields.

From “Who Wins if AI Models Commoditize? — With Mistral CEO Arthur Mensch”, Jan 16, 2026. [Watch the full episode](https://realalpha.app/v/xxUTdyEDpbU/)

## Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

[25:44](https://realalpha.app/v/xxUTdyEDpbU/?t=1544) Well, if you look at the trends in 2024, I'd say there might have been like a six-month gap. If you look at the trend in 2025, I think the gap is more around three months. So I guess it's up to anyone to guess what the gap is going to be next year. But effectively, this gap has been shrinking quite significantly. The reason for that is that basically you have a saturation effect when you pre-train models around 10 to the power of 26 flops.

[26:17](https://realalpha.app/v/xxUTdyEDpbU/?t=1577) the reason for that is that there's only that much data you can find to compress when you pre-train models and so effectively labs that maybe started a little behind created enough compute capacity to train models at this kind of scale and efficiency has also increased and so what it means is that today everybody has access to 10 to 26 flops facilities over the course of a few months. And that's a measure of compute.

[26:46](https://realalpha.app/v/xxUTdyEDpbU/?t=1606) That's a measure of compute. That's a measure of compute times. So you need to tend to 6 flops is something that any lab today can achieve in a couple of months. And because of that, the saturation effect means that open source models have caught up because closed source models that were started had kind of run into that wall of pre-training. so what that means is that this is only going to continue shrinking and if we look at like the latest open release we did which is DevStyle 2 which is a coding model well it's performing I think around the performance of Entropic around 2 or 3 months ago so yeah I think the gap is shrinking and again I think the question is probably not posed in the right way that way because it's also offering very two different distinct value propositions because on one side, this is well-managed

[27:44](https://realalpha.app/v/xxUTdyEDpbU/?t=1664) and you will depend on the provider itself. On the other side, well, it takes a little more effort because you will need to own it more. You will need to learn about how to customize it. You will need to use the right tools for doing so. You will need to maintain its deployment if you choose to deploy it on your own facilities. But at the end, this is creating the leverage you need against close-source providers.

[28:09](https://realalpha.app/v/xxUTdyEDpbU/?t=1689) So the two categories are effectively different. But if you look at the pure performance side, they are definitely converging. You mentioned that there's a saturation effect. So without getting too technical, are the models sort of done with getting better? Let me put it this way. Are AI models going to continue to get better given the fact that they all seem to be hitting saturation? They will get better in more and more specific domains.

[28:38](https://realalpha.app/v/xxUTdyEDpbU/?t=1718) In that, I think we really collectively made them very clever and able to read them about long context and able to call multiple tools, etc. But if you go and want to effectively put them into production in a bank in a manufacturing company, well, the models need to learn about all of the knowledge that is contained into the companies themselves. And so what it effectively means is that for very precise directions, let's say I want to make my model extremely good at discovering materials or extremely good at designing planes, I will need to go and sweat it a little bit and get the right reward signal and get the right experts and ask them to make my model specifically good in that very precise direction. And so we are definitely not done doing that because what we are all racing for is the right environment and the right signal provider for specific capabilities.

[29:37](https://realalpha.app/v/xxUTdyEDpbU/?t=1777) But the broad horizontal reasoning capabilities are still going to improve them, but nobody is going to improve them in a way that is creating a strong gap versus its competitors. So the strong gap is actually in working with vertical experts that know exactly how they design a plane and that actually explain to the model how to do it. And you have like a wealth of directions that you can take because you can do it in physics, you can do it in chemistry, pharmaceutical, in biology.

[30:09](https://realalpha.app/v/xxUTdyEDpbU/?t=1809) And so to me, the most exciting part of what's going to happen in the next two years is that explosion of very precise directions in which the model are going to get better. And for us, the opportunity is to have the right platform for enabling this kind of verticalization, whether with enterprises or you have AI startups, actually, that are working on very verticalized capabilities, and we're happy to help them as well.

[30:35](https://realalpha.app/v/xxUTdyEDpbU/?t=1835) So that's my view of where the field is going to go. We have been about horizontal intelligence growing and things getting more and more clever. and the next two years is going to be about taking models and making them extremely good at a certain skill set and that's actually more exciting because we are getting to a point where if you pick a domain you can just make it superhuman but we are not going to make it superhuman in every domain at the same time Okay, but then on that note earlier in our conversation you said that you're not going to have a model that can do everything but if that training gets done in certain verticals, why not?

[31:15](https://realalpha.app/v/xxUTdyEDpbU/?t=1875) Well, we are also getting to a point where the verticals that you choose do not really transfer to the others. So there's no point in making a model that is good at very precise biology and very precise physics. Because the transfer in between those things is actually pretty unclear. The problem is that if you actually want your model to be able to solve every problem at the same time, you're making it very big, very expensive, and very costly to serve.

[31:38](https://realalpha.app/v/xxUTdyEDpbU/?t=1898) So specialized models is really, you're going to specialize one for bio, one for chemistry, one for like this particular physics problem. Well, it actually makes more sense because if you want to run it at scale, if you want it to run on the background, if you want it to run day and night thinking about specific problems, well, you want it to be as small as possible because the cost of a model is actually proportional to its size.

[31:59](https://realalpha.app/v/xxUTdyEDpbU/?t=1919) And if you inflate the size by making the model great at multiple modes, well, you're actually not very efficient if you want to deploy it and use it as much as possible. So if you look at the economies of it, it does make sense to make specialized models in certain directions. Let me ask you a little bit about the Mistral competitive area. I think that we're here in the US.