Qwen 3.8-27B goes openweight in 2 days | Hacker News Reader

Qwen 3.8-27B goes openweight in 2 days

huggingface.co

19 pointsnateb202222 comments

you can see the countdown here: https://modelscope.cn/models/Qwen/Qwen3.8-2.4T-A95B?spm=a2ty...

That is for 2.5T parameter one, this is the correct countdown for. the 27B one https://modelscope.cn/models/Qwen/Qwen3.8-27B

this seems to be a countdown for the 2.4t model - which is gigantic and is not as exciting for those running models locally

it looks like it says the 27b will be included in the release too, we will find out in 30 minutes lol

Nope 27B is for Friday: https://modelscope.cn/models/Qwen/Qwen3.8-27B

counting down to:

Friday 2026-08-14 17:00 CEST

Friday 2026-08-14 15:00 UTC

Friday 2026-08-14 08:00 PDT

I guess this one is for Qwen3.8-2.4T-A95B from the URL

404 URL that does not give any evidence for the claim made in the post title

It used to feature a countdown, at least when I posted. 2 days and 2 hours from Wednesday 12 August at ~11am. Huggingface or Qwen took down the URL it seems.

Edit: looks like the modelscope.cn URL that other commenters posted also got taken down.

Is that when that URL will work too?

URL was taken down; seems like the modelscope.cn link that other commenters posted also got taken down.

maybe there's some kind of a delay that they have to remove the countdown. I saw it too when it had the countdown and now it's 404

I'm very excited to try this out.

where are you getting that deadline from?

Someone took modelscope countdown URL for 2.4T model and tried 27B path:

https://modelscope.cn/models/Qwen/Qwen3.8-27B

This modelscope.cn url, like my huggingface url, seems to have been taken down.

I love these small models that are able to fit onto local hardware. Even they are capable enough to be helpful and while I think AI is still a bubble, these open models give me faith that the underlying tech isn't going away.

low-quality post from >10,000 karma user lol

As an owner of a company spending $4000/mo on token spend. Yawn...

I tried every model under 1T and they are not useful for AI Agents.

If I want to use a regular chat model about amoral actions, the 7-9B abliterated models were fine.

I'm sure these 27B models can get you the weather after 5 minutes for the people who didn't buy Nvidia chips.

this is a really new account with 6 of Karma, why does the owner of a company spending $4k come to talk about the 27B model useful just get you the weather? companies went south on token usage, and the default for copilot AFAIK is like 40 (or 70 USD)... if we use that to ballpark the company size, it means this person is owner of a company with 400+ it guys, or 100 with 200USD allocation, which is really big... whoever is waiting for the release of the 27B model knows the thing is capable (I'm waiting for it)... it looks so much like a disinformation campaing...

As the part owner of a company using local LLMs we're getting real value out of 27b sized models. I'm saving $1000/mo vs throwing a frontier model at a rather straightforward (for AI) automation need.

The trick is acknowledging they're B or C tier intelligence and speed and using them accordingly.

As a small business you must be really really small to only be at $1000/mo token spend.

Part time/side gig?

Are you new to HN?

Why would you even doubt these claims? Do you use AI Agents or local models?

On the bright side, I think you work for some big company who doesn't use AI Agents at any capacity that we do. Maybe small business will win at our flexibility.

Reply on news.ycombinator.com