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

OpenAI O3-Mini

openai.com

Post reply on HN
Print view View original

111–120 of 944 posts

Re: OpenAI O3-Mini

#111

Post by thot_experiment »

xeckr wrote: post #93
thunkingdeep wrote: post #31

I’ll take the China Deluxe instead, actually. I’ve been incredibly pleased with DeepSeek this past week. Wonderful product, I love seeing its brain when it’s thinking.

Have you tried seeing what happens when you speak to it about topics which are considered politically sensitive in the PRC?

R1 (70B-distill) itself is very uncensored, will give you full account of tiannanmen square from vague prompts. Asking R1 "what significant things happened in china in 1989" had it volunteering that "the death toll was in the hundreds or thousands and the exact number remains disputed to this day". The only thing that's censored is the web interface.

Re: OpenAI O3-Mini

#112

Post by georgewsinger »

anothermathbozo wrote: post #88
georgewsinger wrote: post #57

Did anyone else notice that o3-mini's SWE bench dropped from 61% in the leaked System Card earlier today to 49.3% in this blog post, which puts o3-mini back in line with Claude on real-world coding tasks? Am I missing something?

I think this is with and without "tools." They explain it in the system card: > We evaluate SWE-bench in two settings: > *• Agentless*, which is used for all models except o3-mini (tools). This setting uses the Agentless 1.0 scaffold, and models are given 5 tries to generate a candidate patch. We compute pass@1 by averaging the per-instance pass rates of all samples that generated a valid (i.e., non-empty) patch. If…

Makes sense. Thanks for the correction.

Re: OpenAI O3-Mini

#113

Post by leovander »

rafaquintanilha wrote: post #104
leovander wrote: post #87

Earlier quoted context omitted.

I am running the 7B distilled version locally. I asked it to create a skeleton MEAN project. Everything was great but then it started to generate the front-end and I noticed the file extension (.tsx) and then saw react getting imported. I gave the same prompt to sonnet 3.5 and not a single hiccup. Maybe not an indication that Deepseek is worse/bad (I am using a distilled version), but moreso speaks to much react/next…

You know you are running an extremely nerfed version of the model, right?

I did update my comment, but said that I am using the distilled version, so yes?

Re: OpenAI O3-Mini

#115

Post by ANewFormation »

rf15 wrote: post #37
sho_hn wrote: post #12

Earlier quoted context omitted.

I think OpenAI really needs to rethink its product naming, especially now that they have a portfolio where there's no such clear hierarchy, but they have a place along different axis (speed, cost, reasoning, capabilities, etc). Your summary attempt e.g. also misses o3-mini vs o3-mini-high. Lots of trade-ofs.

They're strongly tied to Microsoft, so confusing branding is to be expected.

I can't wait for Project Unify which just devolves into a brand new p3-mini type naming convention. It's pretty much identical to the o3-mini, except the API is changed just enough to be completely incompatible and it crashes on any query using a word with more than two syllables. Fix coming soon, for 4 years so far.

On the bright side the app now has curved edges!

Re: OpenAI O3-Mini

#116

Post by ryanhecht »

aleph_minus_one wrote: post #85
ryanhecht wrote: post #66

Earlier quoted context omitted.

But what would we call that model?

> But what would we call that model? Ask one of their models for advice. :-)

Reminds me of a joke in the musical "How to Succeed in Business Without Really Trying" (written in 1961):

PETERSON Oh say, Tackaberry, did you get my memo?

TACKABERRY What memo?

PETERSON My memo about memos. We're sending out too many memos and it's got to stop!

TACKABERRY All right. I'll send out a memo.

Re: OpenAI O3-Mini

#117

Post by winrid »

golly_ned wrote: post #82
kingnothing wrote: post #55

Earlier quoted context omitted.

They really need someone in marketing. If the model is for technical stuff, then call it the technical model. How is anyone supposed to know what these model names mean? The only page of theirs attempting to explain this is a total disaster. https://platform.openai.com/docs/models

Yes, this $300Bn company generating +$3.4Bn in revenue needs to hire marketing expert. They can begin by sourcing ideas from us here to save their struggling business from total marketing disaster.

At the least they should care more about UX. I have no idea how to restore the sidebar on chatgpt on desktop lol

Re: OpenAI O3-Mini

#119

Post by thefourthchime »

thunkingdeep wrote: post #31

I’ll take the China Deluxe instead, actually. I’ve been incredibly pleased with DeepSeek this past week. Wonderful product, I love seeing its brain when it’s thinking.

I've seen it get into long 5 minute chains of thought where it gets totally confused.

Re: OpenAI O3-Mini

#120

Post by thot_experiment »

c-fe wrote: post #107
mechagodzilla wrote: post #63

Earlier quoted context omitted.

Being able to see the thinking trace in R1 is so useful, as you can go back and see if it's getting stuck, making a wrong assumption, missing data, etc. To me that makes it materially more useful than the OpenAI reasoning models, which seem impressive, but are much harder to inspect/debug.

I would actually love if it would just ask me simple questions (just yes/no) when its thinking about something i wasnt clear about and i could help it this way, its a bit sad seeing it write out the assumption and then take the wrong conclusion

You can run it locally, pause it when it thinks wrong and correct it's chain of thought.
Post reply on HN