# Can Open-Source Models Distill OpenAI To Death? | Greg Brockman

*Big Technology* · https://www.youtube.com/watch?v=OhBzfRA_oZQ

## Summary

In this short interview, Greg Brockman responds to the argument that open-source model makers can distill OpenAI's leading models and stay only a couple of months behind. He argues OpenAI's real investment is in the "machine that makes the machine" — the end-to-end co-designed system of people and supercomputers — and that distillation cannot fully replicate a model's capability. He also addresses pricing and competition by pointing to steep year-over-year price drops, Jevons paradox, and the returns to intelligence seen in the new model.

## Key points

- Brockman says OpenAI's real investment is in end-to-end co-design: a system of people and a way of leveraging massive supercomputers to produce models.
- He argues distillation is not as simple as taking a model's output and getting the same capability in a smaller, faster package — otherwise OpenAI would do it itself.
- OpenAI invests heavily in safeguards and mitigations, including a long-standing public preparedness framework covering areas like cyber and bio.
- Brockman notes OpenAI has dropped prices on the same level of intelligence year over year, sometimes by a factor of a hundred.
- He describes a real Jevons paradox: lowering the cost of intelligence leads to way more activity.
- He says there are returns to intelligence, and a little more intelligence goes a long way in what people use the models for.
- He claims the new model's "point one" improvement understates the magic early testers have seen in practical work.

## Notes

### The distillation question
- The interviewer describes a repeating pattern: massive models come out, open-source makers distill them, and open source ends up just a couple of months behind the leading foundational models.
- With investment now so large and capabilities increasing dramatically, the interviewer asks how OpenAI's lead is defensible long term.
- Brockman reframes the question: the real investment is in end-to-end co-design — a system of people producing the technology and a way of working together, including how to leverage massive supercomputers.

### Why distillation isn't enough
- Brockman argues it's not as simple as taking a model's outputs and distilling them into an equally capable but smaller, faster model.
- If that were possible, OpenAI would do it itself and gain a model that is much easier to serve.
- He acknowledges there is a lot of art behind distillation, but stresses that OpenAI is investing in "the machine that makes the machine."

### Safety and deployment
- On the deployment side, OpenAI thinks a lot about safeguards and mitigations for many ways the models could be misused in real situations.
- This spans areas like cyber and bio, backed by a long-standing effort documented in the public preparedness framework.
- Brockman says every piece of what OpenAI does must connect to making progress while making the models broadly available, because the company believes the technology empowers and lifts everyone up.

### Pricing and the open-source economics threat
- The interviewer notes the new model's pricing is roughly double that of GPT-5.4, and asks how OpenAI handles open source delivering almost-as-good performance more cheaply.
- Brockman responds that OpenAI's history shows prices on the same level of intelligence dropping year over year, sometimes by a factor of a hundred, and at least an order of magnitude.
- He cites a real Jevons paradox: when you lower the cost of something, way more activity happens.

### Returns to intelligence
- Brockman says there are returns to intelligence — for the tasks these models can now do, a little more intelligence goes a long way.
- He frames the new model (5.5) as seemingly an incremental improvement, but predicts a massive improvement in what people use it for.
- He says the "point one" improvement really undersells the magic that early testers have seen in their practical work.

## Transcript

[00:00] And there's been this pattern where these massive models come out uh they get distilled uh by open source model makers and then open source is just a couple months behind the leading foundational models. Um and you know, I guess like when the investment was smaller, being a couple months ahead, uh you know, mattered a lot. But I'm curious now that the investment is so big um and the models are capabilities are increasing you know, fairly dramatically you know, uh as you go. Um how is this defensible in the long term if you're just going to have that pattern repeat over and over? Well, I look at it a little differently. Like I think that the real investment that we are making is into that end-to-end co-design, right? Of

[00:44] having a system a system of people, right? Who are producing this technology, right? A way of working together and some of this is about how you leverage these massive supercomputers to produce these models. Now, it is also the case that it's not as simple as you can take the output of these models and distill and you have exactly the model of the same capability. It's just smaller and can run fast. If that were the case, we would just do that and then we would also have a model that would be, you know, much more uh easy to serve in in many ways. And of course, there's a lot of art behind distillation. There's a lot of great things there. But the point that I'm getting at is that the real thing that

[01:20] we are investing in is the machine that makes the machine. Now, the at the deployment side, we think a lot about safeguards. We think a lot about mitigations and we do that for many many different aspects of how these models could be misused um in real situations. And that's something that we have been investing in for many years and we think about that across areas like cyber or thinking about that areas like bio that we have a I a long-standing effort that you can see in our preparedness framework, which is public, about how we approach these kinds of uses of the model and how we try to make maximize the benefits, mitigate the risks. And so I think it's a real motion that every piece of what

[01:59] we do needs to connect to the question of how do we continue to make progress, but also how do we make these models broadly available? Because that's something that we really believe in, that we believe this technology empowers people, and that we want it to benefit people and lift everyone up. Yeah, but just to go back on that, um the pricing on this model is, I think, double the last model, GPT-5.4. Um and so, from an economics or a business standpoint, the question would be, you know, let's say you keep on progressing, but because there's been all this infra- infrastructure that's been put towards training the models, if open source uh can deliver not as good performance, but almost as good, uh and do it cheaper, um

[02:42] how do you handle that threat? Well, again, I look at it a little differently. So, first of all, if you look at our history, which really is not driven by anything in competition, it's just like our our own sort of progress and and desire, we have dropped prices on the same level of intelligence year over year, sometimes by literally a factor of a hundred. Right? It's like, at least an order of magnitude year over year, sometimes literally a hundred. But, the thing that keeps happening, it's real Jevons paradox, where it's like you lower the cost of something, way more activity happens, right? And I think that what we keep seeing is that there are returns to intelligence, right? That for the kinds of tasks that

[03:21] these models are now capable of doing, that a little bit more intelligence goes a long way. And I think that is the story of 5.5, that in some ways, you can almost look at it as like, "Oh, there's just an incremental improvement in intelligence." But, I think there's going to be a massive improvement in terms of what people use it for. And by the way, I actually think that incremental is actually very much an understatement for this model relative to 5.5. You know, it's a point one improvement in some ways, um but I think that I that that actually really undersells the magic that we see within this model, and that the that our early testers have have really [music] seen in their practical work.