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

821–830 of 909 posts

Re: Open models by OpenAI

#821

Post by brna-2 »

nialv7 wrote: post #819
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".

This is not actually just about having it produce text that is censored but doing anything it says it is not allowed to do at all. I am sure these two mostly overlap but not always. Like I said, it is not allowed to have "no output" and it is hard to make it do it.

Re: Open models by OpenAI

#822

Post by aaroninsf »

9rx wrote: post #637
int_19h wrote: post #496

Earlier quoted context omitted.

I tried 20b locally and it couldn't reason a way out of a basic river crossing puzzle with labels changed. That is not anywhere near SOTA. In fact it's worse than many local models that can do it, including e.g. QwQ-32b.

I tried the two US presidents having the same parents one, and while it understood the intent, it got caught up in being adamant that Joe Biden won the election in 2024 and anything I do to try and tell it otherwise is dismissed as being false and expresses quite definitely that I need to do proper research with legitimate sources.

Have we considered the possibility that maybe it knows something we don't.

Re: Open models by OpenAI

#823

Post by djeastm »

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.

Sounds like a pre-Beatles "guitar groups are on their way out" kind of statement

Re: Open models by OpenAI

#824

Post by hrpnk »

rt1rz wrote: post #452
hrpnk wrote: post #388

Earlier quoted context omitted.

gpt-oss-20b: 9 threads, 131072 context window, 4 experts - 35-37 tok/s on M2 Max via LM Studio.

interestingly, i am also on M2 Max, and i get ~66 tok/s in LM Studio on M2 Max, with the same 131072. I have full offload to GPU. I also turned on flash attention in advanced settings.

Thank you! Flash attention gives me a boost to ~66 tok/s indeed.

Re: Open models by OpenAI

#825

Post by FrustratedMonky »

9rx wrote: post #804
FrustratedMonky wrote: post #796

Earlier quoted context omitted.

You think the Democratic White House, manipulated Republicans into Voting for Trump. So it is the Democrats fault we have Trump??? Next Level Cope.

> You think the Democratic White House, manipulated Republicans into Voting for Trump. Yes, that is what he thinks. Did you not read the comment? It is, like, uh, right there... He also explained his reasoning: If Trump didn't win the party race, a more compelling option (the so-called "50-year-old youngster") would have instead, which he claims would have guaranteed a Republican win. In other words, what he is sayin…

"explained his reasoning"

Well, I guess, if you are taking some pretty wild speculation as a reasoned explanation. There isn't much hope for you.

Maybe it was because the Democrats new the Earth was about the be invaded by an Alien race , and they also knew Trump was actually a lizard person (native to Earth and thus on their joint side). And Trump would be able to defeat them, so using the secret mind control powers, the Democrats were able to sway the election to allow Trump to win and thus use his advanced Lizard technology to save the planet. Of course, this all happened behind the scenes.

I think if someone is saying the Democrats are so powerful and skillful, that they can sway the election to give Trump the primary win, but then turn around and lose. That does require some clarification.

I'm just hearing a lot of these crazy arguments that somehow everything Trump does is the fault of the Democrats. They are crazy on the face of it. Maybe if people had to clarify their positions they would realize 'oh, yeah, that doesn't make sense'.

Re: Open models by OpenAI

#826

Post by FrustratedMonky »

9rx wrote: post #816
FrustratedMonky wrote: post #754

Earlier quoted context omitted.

[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 i…

I was just referring to the predominant number of cases where Church officials, and Republicans are caught in under-age scandals. It seems like it is coming out of the shadows now, and Republicans are just openly going with it, they like em young and illegal. Epstein is just the case where the 'right' bothered keeping up tabs on it, so now they are clutching their pearls.

Re: Open models by OpenAI

#827

Post by jwr »

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…

gpt-oss:20b is the best performing model on my spam filtering benchmarks (I wrote a despammer that uses an LLM).

These are the simplified results (total percentage of correctly classified E-mails on both spam and ham testing data):

gpt-oss:20b 95.6%

gemma3:27b-it-qat 94.3%

mistral-small3.2:24b-instruct-2506-q4_K_M 93.7%

mistral-small3.2:24b-instruct-2506-q8_0 92.5%

qwen3:32b-q4_K_M 89.2%

qwen3:30b-a3b-q4_K_M 87.9%

gemma3n:e4b-it-q4_K_M 84.9%

deepseek-r1:8b 75.2%

qwen3:30b-a3b-instruct-2507-q4_K_M 73.0%

I'm quite happy, because it's also smaller and faster than gemma3.

Re: Open models by OpenAI

#828

Post by CMay »

aspect0545 wrote: post #718
CMay wrote: post #699

Earlier quoted context omitted.

The 20b solved the wolf, goat, cabbage river crossing puzzle set to high reasoning for me without needing to use a system prompt that encourages critical thinking. It managed it using multiple different recommended settings, from temperatures of 0.6 up to 1.0, etc. Other models have generally failed that without a system prompt that encourages rigorous thinking. Each of the reasoning settings may very well have think…

But was it reasoning or did it solve this because it was parting it‘s training data?

Maybe both? I tried using different animals, scenarios, solvable versions, unsolvable versions, it gave me the correct answer with high reasoning in LM Studio. It does tell me it's in the training data, but it does reason through things fairly well. It doesn't feel like it's just reciting the solution and picks up on nuances around the variations.

If I switch from LM Studio to Ollama and run it using the CLI without changing anything, it will fail and it's harder to set the reasoning amount. If I use the Ollama UI, it seems to do a lot less reasoning. Not sure the Ollama UI has an option anywhere to adjust the system prompt so I can set the reasoning to high. In LM Studio even with the Unsloth GGUF, I can set the reasoning to high in the system prompt even though LM Studio won't give you the reasoning amount button to choose it with on that version.

Re: Open models by OpenAI

#829

Post by 9rx »

FrustratedMonky wrote: post #826
9rx wrote: post #816

Earlier quoted context omitted.

> 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 i…

I was just referring to the predominant number of cases where Church officials, and Republicans are caught in under-age scandals. It seems like it is coming out of the shadows now, and Republicans are just openly going with it, they like em young and illegal. Epstein is just the case where the 'right' bothered keeping up tabs on it, so now they are clutching their pearls.

> I was just referring to the predominant number of cases where Church officials, and Republicans are caught in under-age scandals.

But even that is characterized by the "choir boy", not the "baby being baptized". Where is this pedophilia idea coming from?

Re: Open models by OpenAI

#830

Post by canvascritic »

nojito wrote: post #535
canvascritic wrote: post #404

Earlier quoted context omitted.

Healthcare organizations that can't (easily) send data over the wire while remaining in compliance Organizations operating in high stakes environments Organizations with restrictive IT policies To name just a few -- well, the first two are special cases of the last one RE your hallucination concerns: the issue is overly broad ambitions. Local LLMs are not general purpose -- if what you want is local ChatGPT, you will…

Pretty much all the large players in healthcare (provider and payer) have model access (OpenAI, Gemini, Anthropic)

This may be true for some large players in coastal states but definitely not true in general

Your typical non-coastal state run health system does not have model access outside of people using their own unsanctioned/personal ChatGPT/Claude accounts. In particular even if you have model access, you won't automatically have API access. Maybe you have a request for an API key in security review or in the queue of some committee that will get to it in 6 months. This is the reality for my local health system. Local models have been a massive boon in the way of enabling this kind of powerful automation at a fraction of the cost without having to endure the usual process needed to send data over the wire to a third party

Post reply on HN