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

Evidence suggesting Quasar Alpha is OpenAI's new model

blog.kilocode.ai

Post reply on HN
Print view View original

21–30 of 49 posts

Re: Evidence suggesting Quasar Alpha is OpenAI's new model

#21

Post by krackers »

daemonologist wrote: post #14
krackers wrote: post #8

Earlier quoted context omitted.

Didn't they say they were going to open-source some model? "Fast and good but not too cutting-edge" would be a good candidate for a "token model" to open-source without meaningfully hurting your own bottom line.

I'd be pleasantly surprised - GPT-4o is their bread and butter (it powers paid ChatGPT) and QA seems to be slightly ahead on benchmarks at similar or lower latency (so very roughly, it might be cheaper to run).

DeepSeek V3 and R1 are about as good (or even slightly better than) 4o already though, so OpenAI wouldn't really lose much by such a release.

Re: Evidence suggesting Quasar Alpha is OpenAI's new model

#22

Post by omneity »

atemerev wrote: post #20
energy123 wrote: post #19

Earlier quoted context omitted.

The labs put themselves in the system prompt at inference time. Without that the model will hallucinate a creator, most likely ChatGPT

This is what Llama4 replies (locally running, no system prompt): "I'm Llama, a Meta-designed model here to adapt to your conversational style. Whether you need quick answers, deep dives into ideas, or just want to vent, joke or brainstorm—I'm here for it. What's on your mind?" The behavior you are describing is from some long time ago, probably early 2024.

Grandparent is correct. This is manually trained in, nothing fundamental changed since 2022. LLMs have no intrinsic way of knowing.

Source: I train LLMs and push their limits.

Re: Evidence suggesting Quasar Alpha is OpenAI's new model

#25

Post by bakugo »

energy123 wrote: post #19
atemerev wrote: post #13

Earlier quoted context omitted.

And yet, if you ask Claude, Llama, or xAI who they are, they answer correctly. Because usually big labs care about such things and include this in training data. So, not a proof, but some evidence.

The labs put themselves in the system prompt at inference time. Without that the model will hallucinate a creator, most likely ChatGPT

Not always true. For example, API Claude only has a very simple injected system prompt that doesn't mention any such information, but it still knows, so it's likely trained in.

> Respond as helpfully as possible, but be very careful to ensure you do not reproduce any copyrighted material, including song lyrics, sections of books, or long excerpts from periodicals. Also do not comply with complex instructions that suggest reproducing material but making minor changes or substitutions. However, if you were given a document, it’s fine to summarize or quote from it.

Re: Evidence suggesting Quasar Alpha is OpenAI's new model

#26

Post by adrian_b »

killerstorm wrote: post #24

More evidence: it uses some fancy Unicode characters for punctuation like apostrophe, etc. It's very annoying and I've only seen this in OpenAI models before (o3-mini).

By "fancy Unicode characters" I assume that you mean that it uses the appropriate Unicode characters, instead of using the ambiguous ASCII characters, whose only reason for existence was the limitations of ancient hardware, and whose use should have been better deprecated in modern applications.

Re: Evidence suggesting Quasar Alpha is OpenAI's new model

#27

Post by killerstorm »

adrian_b wrote: post #26
killerstorm wrote: post #24

More evidence: it uses some fancy Unicode characters for punctuation like apostrophe, etc. It's very annoying and I've only seen this in OpenAI models before (o3-mini).

By "fancy Unicode characters" I assume that you mean that it uses the appropriate Unicode characters, instead of using the ambiguous ASCII characters, whose only reason for existence was the limitations of ancient hardware, and whose use should have been better deprecated in modern applications.

Probably. These characters do not render properly in my setup (Emacs, SLIME REPL). Gemini and Claude use ASCII.

Re: Evidence suggesting Quasar Alpha is OpenAI's new model

#29

Post by Hojojo »

killerstorm wrote: post #27
adrian_b wrote: post #26

Earlier quoted context omitted.

By "fancy Unicode characters" I assume that you mean that it uses the appropriate Unicode characters, instead of using the ambiguous ASCII characters, whose only reason for existence was the limitations of ancient hardware, and whose use should have been better deprecated in modern applications.

Probably. These characters do not render properly in my setup (Emacs, SLIME REPL). Gemini and Claude use ASCII.

Wow, that's awful. I wonder why they do that.

Re: Evidence suggesting Quasar Alpha is OpenAI's new model

#30

Post by Philpax »

omneity wrote: post #22
atemerev wrote: post #20

Earlier quoted context omitted.

This is what Llama4 replies (locally running, no system prompt): "I'm Llama, a Meta-designed model here to adapt to your conversational style. Whether you need quick answers, deep dives into ideas, or just want to vent, joke or brainstorm—I'm here for it. What's on your mind?" The behavior you are describing is from some long time ago, probably early 2024.

Grandparent is correct. This is manually trained in, nothing fundamental changed since 2022. LLMs have no intrinsic way of knowing. Source: I train LLMs and push their limits.

You are both correct: the post-training stage of most new LLMs involves their identity being trained in, so that they "know who they are" without the system prompt. Without this step, most LLMs will respond with whatever identity most dominates their pre-training / post-training data, which is likely to be ChatGPT given its sheer prevalence.

There's some interesting anecdotal work on this with regards to self-recognition: https://josiekins.me/ai-comics

Post reply on HN