Astro Hacker News - Qwen3.8 27B

bravetraveler |next [-]

Didn't think we'd have model launch pages so 'soon'; how long until we have something Steam-like and can preload our eagerly-awaited models?

Anyway, eager to try [on hardware I own]!

Gecko4072 |next |previous [-]

DeepSwe score is 42.2. For comparison 3.6-27b is 13.3, GLM 5.2 is 44, and Opus 4.8 is 59.

meffmadd |next |previous [-]

Hopefully they release the 35B-A3B version alongside it.

wolvoleo |next |previous [-]

I miss a good 9B class model :',( The last one was 3.5.

27B is just a little bit too big for a 16GB GPU.

nezhar |root |parent [-]

Yeah, same with the latest releases of meta and NVIDIA. It's like everybody is expected to have 32 GB of RAM

|next |previous [-]

WithinReason |next |previous [-]

Trading blows with Opus 4.6 is impressive, awaiting unsloth quants

nezhar |next |previous [-]

It feels so strange that we now have countdowns for model releases

jimmydoe |root |parent [-]

Alibaba does this for their shopping biz all the time. Growth hack is very high priority in the company. Their homepage was a bunch of ridiculous SEO. The whole culture there seems very distasteful to me. This doesn’t mean this qwen model is bad today, but I’m a believer that culture will decide the product quality in the long run.

ChrisArchitect |next |previous [-]

Later discussion: https://news.ycombinator.com/item?id=49299684

darthoctopus |next |previous [-]

this page gives a 404

kristjansson |root |parent [-]

Oh shoot it had a nice countdown timer to ~8a PT when I submitted it

breakingcups |root |parent [-]

It's back for me

jimmydoe |previous [-]

-