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

Hy3

hy.tencent.com

Post reply on HN
Print view View original

61–70 of 125 posts

Re: Hy3

#61

Post by andai »

SwellJoe wrote: post #46
thot_experiment wrote: post #35

I feel like I'm taking crazy pills with hy3, it's either benchmaxxed to hell and back or skill issue on my part but I'd rather use dense gemma. I don't think there's a single model that's wasted more of my time in recent memory.

The Hy3 preview has been a mediocre performer in my benchmarks of security auditing with models, and yes, it is outperformed by Gemma 4 (31b soundly beats it, the MoE does slightly better, even at 4-bit quantization when using the QAT version). Qwen 3.6 27b also beats it. I'll try it again now that it's out of preview and has been updated with more post-training. It presumably can't be worse, so maybe it's better eno…

For context Hy is 295B, roughly on par with DeepSeek V4 Flash (284B).

I haven't tested it yet so I cannot comment on the quality (nor the comparison with 10x smaller (!) models)

Re: Hy3

#62

Post by thot_experiment »

andai wrote: post #61
SwellJoe wrote: post #46

Earlier quoted context omitted.

The Hy3 preview has been a mediocre performer in my benchmarks of security auditing with models, and yes, it is outperformed by Gemma 4 (31b soundly beats it, the MoE does slightly better, even at 4-bit quantization when using the QAT version). Qwen 3.6 27b also beats it. I'll try it again now that it's out of preview and has been updated with more post-training. It presumably can't be worse, so maybe it's better eno…

For context Hy is 295B, roughly on par with DeepSeek V4 Flash (284B). I haven't tested it yet so I cannot comment on the quality (nor the comparison with 10x smaller (!) models)

There are extreme diminishing returns in real world performance as models get bigger. 10x bigger might mean 5-10% better on benchmarks, a margin that can easily mean it's functionally equivalent in real world use or even a worse performer depending on the context it's being used in, and how good you are at providing meaningful context.

Of course the bigger model embeds more knowledge, but when neither model has the knowledge necessary to perform the task, hy3 makes idiotic decisions all the time whereas gemma 31b has a decent hit rate.

hy3 feels like someone who's read a lot of books and says the right words but has nothing of substance between their ears, gemma feels like a reasonably intelligent person who doesn't understand the domain, the latter is muuuch easier to work with than the former.

Re: Hy3

#64

Post by jszymborski »

I'm sorry but what on earth is going on with that bar chart, the bars are not consistent. E.g., in the frontierscience-olympiad chart Hy3 preview scores the same as DeepSeek (70.0) but Hy3 preview's bar is visibly lower.

Re: Hy3

#65

Post by SwellJoe »

thot_experiment wrote: post #62
andai wrote: post #61

Earlier quoted context omitted.

For context Hy is 295B, roughly on par with DeepSeek V4 Flash (284B). I haven't tested it yet so I cannot comment on the quality (nor the comparison with 10x smaller (!) models)

There are extreme diminishing returns in real world performance as models get bigger. 10x bigger might mean 5-10% better on benchmarks, a margin that can easily mean it's functionally equivalent in real world use or even a worse performer depending on the context it's being used in, and how good you are at providing meaningful context. Of course the bigger model embeds more knowledge, but when neither model has the k…

Gemma 4 is the first really small model that feels smart, to me. I mean, Qwen 3.6 is arguably better at some coding tasks. But, Gemma 4 has shockingly good reasoning for a small model. Even the tiny 12B, at 7GB on disk in the 4-bit QAT quant, feels like a really big model of a couple years ago. It's a good tool user, can search the web (when given the appropriate skill or MCP), has good vision capabilities, and pretty good prose.

I've only used the Hy3 preview, so I don't want to judge too harshly, yet. But, I wasn't very impressed with it a couple of months ago.

Re: Hy3

#66

Post by Azantys »

BoorishBears wrote: post #58
minimaxir wrote: post #24

A month ago I wrote a blog post about how Hy3 was topping the OpenRouter rankings despite no one talking about it: https://news.ycombinator.com/item?id=48317294 As of today, it has fallen to 8/9th on the rankings. I don't see a reason where you would use this model over competitors. However, price economics are bit confusing, as currently the effective input price of Hy3 via OpenRouter is now the same as DeepSeek-hos…

Writes pretty engaging prose, finetunes well, now MIT licensed... what's not to like? Oh and very good world knowledge for the size: better than than DS4 Flash

Do people really use 100B+ models for writing? I am no writer but to me it seems like writing is one of the easiest tasks with barely any logic or reasoning and as long as its not longer than a handful of pages I expect even 8B models to perform great.

Re: Hy3

#67

Post by SwellJoe »

SwellJoe wrote: post #48
james2doyle wrote: post #13

Been using this and GLM 5.2 back and forth. I like the speed of Hy3. Also seems very happy to follow instructions. Still haven’t found any open models that follow instructions as good as Mimo v2 pro though

MiMo v2.5 Pro is very spiky, in my experience. Sometimes excellent, sometimes mediocre. Weirdly high non-deterministic behavior. Run the same task three times, get three different results. I mean, they're all rolling dice for the next word, but MiMo seems to run hot on the randomness dimension in my benchmarks. But, it performs very well for its size. I just looked it up, and it's much smaller than I thought it was w…

Er, actually, Pro is a big 1T model (which makes more sense given how well it does, sometimes). Regular MiMo is small, and I haven't tested it.

Re: Hy3

#68

Post by assimpleaspossi »

andai wrote: post #59
assimpleaspossi wrote: post #51

I would never use any product that can't explain on its own front page what it is and why I should use it.

It's like a computer, but you can talk to it.

If they can't tell me what it is or how to use it then they can't explain to me how to install it or anything else properly so forget it. I'm not wasting my time trying to figure it out.

Re: Hy3

#69

Post by girvo »

UncleOxidant wrote: post #16
Catloafdev wrote: post #9

Curious how people feel about this compared to DS4 Flash, given they are pretty close in size. Also curious how well it holds up to heavy quantization. DS4 Flash can currently run reasonably well on systems with ~96gb+ RAM, I wonder if Hy3 can compete there.

That's a 2-bit quant of DS4 flash. You're probably better off running Qwen3.6-27B at Q8.

Having heavily evaluated both antirez’s ds4 flash and Qwen 3.6 27B at FP8 and Q8: it depends. The quantised Flash is better in a number of tasks despite running much slower on my DGX Spark-alike.

27B is amazing for its size but has some surprising limits when used for longer agentic coding sessions, especially if you’re using tools that are outside the stock standard web tech stuff: it really isn’t good at Relay, for example.

Re: Hy3

#70

Post by VulgarExigency »

wgd wrote: post #34
andai wrote: post #10

Earlier quoted context omitted.

I've been wondering about that. GLM-5.2 is also half the size of DeepSeek V4 Pro. (But costs roughly twice as much.) I looked into DeepSeek's architecture a little bit and the main focus was how can we save as much money as possible. They did a lot of cost cutting with the attention mechanisms. This allowed them to offer an insanely cheap price even on massive contexts, but seems to have come at the cost of performan…

The current Deepseek V4 Pro is still just their initial preview AFAIK, with the "real" model release rumored to come later this month. GLM-5.2 might be outperforming simply because it's had more post-training on top of the GLM-5 base.

If the "final" release of Deepseek V4 Pro outperforms GLM-5.2 while maintaining the current Deepseek price, then it's going to be a marvel.
Post reply on HN