Sounds like an open GPT-4o / o1 distill or something… And that they want to know what to go for.

openAI: we would like your opinions on how an open model would be useful for you!

HN commenters: omg why are you asking for my opinion? shady as fuck

The original comment says "their mentality about open-model is wrong, evident from their post". Does not say it's shady! Why to be cynical?

I'm surprised they expect people to voluntarily undergo that interrogation, for their feedback to then be processed by some LLM.

If you're interested in the field, why wouldn't you answer the questionnaire? It costs you basically nothing while the upside is getting something that is potentially at least a tiny bit more useful to you than if you hadn't said anything?

It takes time and will likely be ignored.

Why would it be ignored? They're the ones asking.

would you write a response just to be read by an LLM?

Sure, I do it everyday when I chat with AIs. Humans not reading our exchanges doesn't bother me particularly.

The idea that they want to train a new custom model for open release instead of just... giving us GPT-3 already suggests a terrible start. I'm calling it now, this is a strategic counterplay against Google's Gemma models so @sama can sell Tim Cook a "frontier" local model that doesn't compete with anything coherent. A fig leaf for their "Open" identity and a paper tiger for the Apple Intelligence panoply.

OpenAI doesn't believe in Open Source, they merely want it's prestige without committing to it on-principle.

> instead of just... giving us GPT-3

If you're referring to the GPT-3 from 2020, modern open source models five years later are a) better at benchmarks b) much smaller yet still better at said benchmarks c) much, much cheaper/faster due to architectural improvements.

The real hard thing for OpenAI to do is to release an open-weights model that's better/more differentiated than Gemma 3 (at the small scale) or DeepSeek R1 (at the large scale)

But if OpenAI was Open, they'd open-source those old obsolete models. You're right that no one really wants it, and it has little to no commercial value at this point, but that's all the more reason they should just put it out there and actually live up to their name.

Giving them the benefit of the doubt: open models are often run on more modest hardware than commercial behemoths like GPT-4. Maybe they want to know what kind of hardware they'd like to run it on? Even the severely outdated 175B GPT-3.5 is still too expensive to run on most machines. We don’t know how unoptimized their models are. Releasing an outdated, unoptimized model is pointless, when much, much better alternatives are already available.

Is there anything about training a new model? I just assumed they were asking for all the little bits that companies forget to do at release with an open model.