Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Open models by OpenAI

openai.com

Post reply on HN
Print view View original

811–820 of 909 posts

Re: Open models by OpenAI

#811

Post by kmacdough »

dongobread wrote: post #433
cco wrote: post #357

The lede is being missed imo. gpt-oss:20b is a top ten model (on MMLU (right behind Gemini-2.5-Pro) and I just ran it locally on my Macbook Air M3 from last year. I've been experimenting with a lot of local models, both on my laptop and on my phone (Pixel 9 Pro), and I figured we'd be here in a year or two. But no, we're here today. A basically frontier model, running for the cost of electricity (free with a rounding…

How up to date are you on current open weights models? After playing around with it for a few hours I find it to be nowhere near as good as Qwen3-30B-A3B. The world knowledge is severely lacking in particular.

I too am skeptical of these models, but it's a reasoning focused model. As a result this isn't a very appropriate benchmark.

Small models are going to be particularly poor when used outside of their intended purpose. They have to omit something.

Re: Open models by OpenAI

#812

Post by collingreen »

FrustratedMonky wrote: post #754
cpursley wrote: post #733

Earlier quoted context omitted.

You’re in a bubble. It was no surprise to folks who touch grass on the regular.

[flagged]

I've been surprised by a lot of the values that have been put aside for the elections recently, especially when so many of them were yelled about so loudly in the past few elections (not limited to one side but from my perspective one side does seem more severe).

It's also been really interesting to watch a bunch of formerly non partisan values become polarizing like rule of law, integrity, corruption, decorum, body autonomy, the constitution, the national debt, global alliances, sexual assault, "family values", and "the American Dream" tm.

There has always been this (baffling, to me) concept that most folks vote for a candidate "while holding their nose" but it seems like the proportions are getting wildly out of whack. The pile of bad smells grows tall and the exchange rate seems quite bad.

The 2020s have taught me a lot about the reality of civics and what it looks like to try to govern a giant group of strangers.

I'm grateful to have lived in the US during a period of stability and safety and (some) shared values. My optimism for retaining that and preventing decline, however, is running very dry.

Re: Open models by OpenAI

#813

Post by organsnyder »

xyc wrote: post #513
organsnyder wrote: post #155

Earlier quoted context omitted.

I read that as it runs in data centers (H100 GPUs) or high-end desktops/laptops (Strix Halo?).

I'm running it with ROG Flow Z13 128GB Strix Halo and getting 50 tok/s for 20B model and 12 tok/s for 120B model. I'd say it's pretty usable.

Excellent! I have a Framework Desktop with 128GB on preorder—really looking forward to getting it.

Re: Open models by OpenAI

#814

Post by liuliu »

throwdbaaway wrote: post #446
liuliu wrote: post #192

Earlier quoted context omitted.

Kimi K2 paper said that the model sparsity scales up with parameters pretty well (MoE sparsity scaling law, as they call, basically calling Llama 4 MoE "done wrong"). Hence K2 has 128:1 sparsity.

I thought Kimi K2 uses 8 active experts out of 384? Sparsity should be 48:1. Indeed Llama4 Maverick is the only one that has 128:1 sparsity.

You are right. I mis-remembered the sparsity part of K2. The "done wrong" part I was thinking about how the scout -> maverick -> behemoth doesn't scale sparsity according to any formula (less sparse -> sparse -> less sparse).

Re: Open models by OpenAI

#815

Post by danielvaughn »

okasaki wrote: post #716
danielvaughn wrote: post #557

Earlier quoted context omitted.

Yep that’s my biggest ask tbh. I just imagine the next Elder Scrolls taking advantage of that. Would change the gaming landscape overnight.

Games with LLM characters have been done and it turns out this is a shit idea.

I guarantee anything that’s already been put out is too early, and is very likely a rushed cash-grab. Which, of course that sucks.

And AI has been in games for a long time. Generated terrain and other sorts of automation have been used as techniques for a hot minute now.

All I’m suggesting is to keep on that same trajectory, now just using an on-device LLM to back intelligence features.

Re: Open models by OpenAI

#816

Post by 9rx »

FrustratedMonky wrote: post #754
cpursley wrote: post #733

Earlier quoted context omitted.

You’re in a bubble. It was no surprise to folks who touch grass on the regular.

[flagged]

> it is a surprise how many people in the country are supporters of pedophilia.

Do you mean ephebophilia? There is no prominent pedophilia movement. The Epstein saga, which is presumably at least somewhat related to what you are referring to, is clearly centred around "almost adults". Assuming that is what you meant, I don't see what is surprising about it. A revolt to the "Teen Mom", "16 and Pregnant" movement was inevitable. People are increasingly growning tired of being shamed for having children.

Re: Open models by OpenAI

#817

Post by matznerd »

matznerd wrote: post #267

thanks openai for being open ;) Surprised there are no official MLX versions and only one mention of MLX in this thread. MLX basically converst the models to take advntage of mac unified memory for 2-5x increase in power, enabling macs to run what would otherwise take expensive gpus (within limits). So FYI to any one on mac, the easiest way to run these models right now is using LM Studio ( https://lmstudio.ai/ ), it…

Here are the LM Studio MLX models:

LM Studio community: 20b: bhttps://huggingface.co/lmstudio-community/gpt-oss-20b-MLX-8b... 120b: https://huggingface.co/lmstudio-community/gpt-oss-120b-MLX-8...

Re: Open models by OpenAI

#818

Post by brna-2 »

Is it just me or is this MUCH sturdier against jailbreaks then similar models, or even the ChatGPT ones?

I have had problems even making it output nothing. But I guess I'll try some more :D

Nice job @openAI team.

Re: Open models by OpenAI

#819

Post by nialv7 »

brna-2 wrote: post #818

Is it just me or is this MUCH sturdier against jailbreaks then similar models, or even the ChatGPT ones? I have had problems even making it output nothing. But I guess I'll try some more :D Nice job @openAI team.

thoughts in the field say instead of a model that is pre-trained normally then censored, this is a model pre-trained on filtered data. i.e. it have never seen anything that is unsafe, ever.

you can't jailbreak when there is nothing "outside".

Re: Open models by OpenAI

#820

Post by sdenton4 »

dns_snek wrote: post #681
sdenton4 wrote: post #352

Earlier quoted context omitted.

I think getting out of the binary is good for the long run. We have something which is artificial, intelligent, and general in scope. We're there. Is it perfect? No. Is it even good? Sometimes! Do airplanes flap their wings? Also no, but they do a lot of stuff nonetheless.

That's where we disagree, I do not consider a system that isn't capable of learning, improving, or reasoning to be generally intelligent. My most basic criteria for "AGI" is a system that can absorb and integrate new knowledge through repetition and experience in real time, just like a human would. Further, their statements, knowledge, and "beliefs" should be reasonably self-consistent. That's where I'm usually told…

Yeah, this is in 'flies like a plane, not like a bird' territory. But I think it's closer than you think.

The systems do learn and have improved rapidly over the last year. Humans have two learning modes - short-term in-context learning, and then longer-term learning that occurs with practice and across sleep cycles. In particular, humans tend to suck at new tasks until they've gotten in some practice and then slept on it (unless the new task is a minor deviation from a task they are already familiar with).

This is true for LLM's as well. They have some ability to adapt to the context of the current conversation, but don't perform model weight updates at this stage. Weight updates happen over a longer period, as pre-training and fine-tuning data are updated. That longer-phase training is where we get the integration of new knowledge through repetition.

In terms of reasoning, what we've got now is somewhere between a small child and a math prodigy, apparently, depending how much cash you're willing to burn on the results. But a small child is still a human.

Post reply on HN