Hy3 • Hacker News Board

Re: Hy3

#81

Post by bjconlan » Thu, Jul 09, 2026, 11:32 PM UTC

verdverm wrote: post #42

spmurrayzzz wrote: post #21

Earlier quoted context omitted.

I think its good advice to test both on your own evals for sure, but the MoE parameters are already natively FP4 in ds4. Dropping to 2bpw isn't as big of a loss as it seems (and as corroborated by antirez's work). Its also only 13B active, so your decode speed would be nearly 2x that of Qwen3.6-27B. So there are other latent benefits as well.

z-lab has been dropping dflash addons for a lot of models https://huggingface.co/collections/z-lab/dflash I'm running the qwen3.6-27B + dflash on a spark and tgen is way up, but keep the draft count low, acceptance rate is terrible beyond half a dozen and it requires more memory

Wow thanks for pointing this out! This is actually what I was hoping would happen when the deepspec stuff dropped! And having zlab create these confirms my bias that I think open models are the way.

Re: Hy3

#82

Post by BoorishBears » Fri, Jul 10, 2026, 12:33 AM UTC

fc417fc802 wrote: post #80

BoorishBears wrote: post #77

Earlier quoted context omitted.

Did you try reading the whole comment? Once creativity is being measured in isolation, getting multiple responses from the model is enough to measure creativity a ton of different ways: wordfreq to identify overused phrases, getting multiple responses for the same prompt and promoting the least similar as preferred for policy optimization, etc. But that's of limited use for stuff like getting diverse names and such.…

I did read the the full comment and I did in fact mean exactly what I wrote when I used the term "unsupervised". I think the condescension does nothing but get in the way. Try extending the benefit of the doubt. > enough to measure creativity a ton of different ways ... The things you listed seem more like temperature than creativity to me. At this point it occurs to me that this is likely yet another case of highly…

The first thing I read from you was a sardonic browbeating in response to the exact comment I gave an earnest response.

And even in domains that lean heavily on "usual phrasing", like technical writing, human writing has notably higher perplexity compared to another LLM's outputs: https://www.sciencedirect.com/science/article/abs/pii/S10766...

With such a low baseline for what's unusual, you do need to get the LLM writing unusual phrases relative to its baseline. Otherwise you get things like repeated n-grams and overused constructs ("it's not X it's Y"), and suddenly the output is predictably not perceived as creative by humans even if you were to insert some otherwise creative or novel premise.

Getting the model to break out of that baseline without disrupting the model's ability to follow technical rules, maintain logic and reasoning, etc. is the difficult part.

-

Also you're again saying unsupervised then following up with descriptions that sure sound like you're referring to RL and supervised learning respectively this time. (supervised learning can improve creativity by the way

Re: Hy3

#83

Post by nmfisher » Fri, Jul 10, 2026, 12:38 AM UTC

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.

I was playing with Hy3 via openrouter yesterday (and I've also been using DS4 Flash/Pro as a daily driver since I cancelled my Anthropic sub a week ago).

I've found DS4 Flash to be very temperental (via Claude Code). The speed is great, but it often builds a completely wrong mental model and charges off down the wrong path. I find myself needing to rein it in regularly (and also compact the history, which undercuts the whole cache price advantage).

Hy3 isn't as fast, but so far it seems to stay on track much more reliably than DS4 Flash. It also doesn't seem to degrade as much with longer context. I'm not sure what the real pricing is, but I feel like it's a very competitive model.

As an aside, I also nabbed a 50m token pack for LongCat 2.0 to give it a whirl. Not free, but it's so cheap they're basically giving it away. Very impressed too - seems roughly on par with Hy3. Not frontier-level intelligence, but a dependable workhorse that can navigate a codebase well and can reliably execute what you tell it to do.

Re: Hy3

#84

Post by invictati » Fri, Jul 10, 2026, 1:06 AM UTC

righthand wrote: post #43

simonw wrote: post #33

Pelican from a few days ago: https://simonwillison.net/2026/Jul/6/hy3/ - I was using the free tier on OpenRouter, which expires on July 21st. I tried the preview model 41 days ago and got a pelican with a "change pelican color" button: https://static.simonwillison.net/static/2026/hy3-preview-pel...

[flagged]

Some things are better left unsaid.

Re: Hy3

#85

Post by mattstir » Fri, Jul 10, 2026, 1:19 AM UTC

This is admittedly an aside from the content of the post itself, but... why do so many mobile sites insist on preventing zooming in and seem to share the same incredibly buggy image zooming? It's quite frustrating.

Re: Hy3

#86

Post by kordlessagain » Fri, Jul 10, 2026, 1:29 AM UTC

righthand wrote: post #43

simonw wrote: post #33

Pelican from a few days ago: https://simonwillison.net/2026/Jul/6/hy3/ - I was using the free tier on OpenRouter, which expires on July 21st. I tried the preview model 41 days ago and got a pelican with a "change pelican color" button: https://static.simonwillison.net/static/2026/hy3-preview-pel...

[flagged]

I will vouch for Simon. I do not know him personally. To me his posts are honest and inspirational.

> I have been overly critical and arguing in bad faith about your writing in the past

I think you are just critical without a stated valid reason. Arguing in bad faith seems to be a thing if HN history is the judge.

And this is coming from a critical thinker, who is a bit tired of people firing off "human slop" comments. The Internet is full of a lot of people, but even when a few are bad apples, it spoils the lot. Maybe that is the intent. Maybe you are just grumpy for your life's situation.

It is 100% possible to build software entirely with AI. If you don't do that, that's great! I still code by hand from time to time, and I'm reading a lot of Rust nowadays, and learning the ropes. I come from a strong Python and Javascript background, plus networking and operations, which I'm a whiz at. I don't do it anymore, but I know how to inform it is done properly.

With this power, I can build things nobody wants to build, but me. Doesn't mean one has to put it into production, or it has to pass some security test, although with me driving it probably will. It only need be what is important to the end user, the prompter, to matter.

I think Simon helps people with this mindset be better at what they love to do. And for that, we should all be grateful.

Re: Hy3

#87

Post by jubilanti » Fri, Jul 10, 2026, 1:29 AM UTC

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…

Because it was free with generous limits and high availability, until it wasn't.

Re: Hy3

#88

Post by kgeist » Fri, Jul 10, 2026, 1:57 AM UTC

ckocagil wrote: post #36

UncleOxidant wrote: post #16

Earlier quoted context omitted.

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

Isn't Q8 way overkill these days? I see many graphs showing Q4 or Q5 having less than %1 deviation. Nvidia's NVFP4 Qwen quantization should be even better due to its better training methods.

Qwen3.6 below Q8 often can't exit a reasoning loop (until it hits max output token count), forgets to insert a tool call, often mistakenly inserts them inside the thinking block... It's still usable though.

Re: Hy3

#89

Post by nullc » Fri, Jul 10, 2026, 1:59 AM UTC

ckocagil wrote: post #36

UncleOxidant wrote: post #16

Earlier quoted context omitted.

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

Isn't Q8 way overkill these days? I see many graphs showing Q4 or Q5 having less than %1 deviation. Nvidia's NVFP4 Qwen quantization should be even better due to its better training methods.

Careful with those graphs, they're usually evaluating the model on KLD on relatively short transcripts. When you're running with 100k token contexts and the model running close loop a difference that looks small in terms of KLD may be quite substantial.

I'm not aware of any great benchmarks that work by giving it a live agentic harness and a number of realistic tasks that take most of the context window to accomplish and evaluate success rate and tokens to completion... but that's what you'd really want to use to judge different quantization levels.

Re: Hy3

#90

Post by dragonwriter » Fri, Jul 10, 2026, 2:01 AM UTC

Azantys wrote: post #66

BoorishBears wrote: post #58

Earlier quoted context omitted.

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.

> I am no writer but to me it seems like writing is one of the easiest tasks with barely any logic or reasoning

Virtually all logic or reasoning is, in one way or another, part of the support for writing. It’s what separates actual writing from generating nonsense that happens to fit grammar rules.

The specific details depend on the domain, of course, but I can’t see how anyone familiar with the output of writing can think that there is little logic or reasoning in doing it well.