this post was submitted on 20 Sep 2026
32 points (82.0% liked)

Selfhosted

62211 readers
882 users here now

A place to share alternatives to popular online services that can be self-hosted without giving up privacy or locking you into a service you don't control.

Rules:

Detailed Rules Post

  1. Be civil.

  2. No spam.

  3. Posts are to be related to self-hosting.

  4. Don't duplicate the full text of your blog or readme if you're providing a link.

  5. Submission headline should match the article title.

  6. No trolling.

  7. Promotion posts require active participation, with an account that is at least 30 days old. F/LOSS without a paywall has exceptions, with requirements. See the rules link for details. Tags [CBH] or [AIP] are required, see the links in Rule 8 for details.

  8. AI-related discussions and AI-involved promotional posts have additional requirements for tagging, as noted in Rule 7 and the AI & Promotional Post Expanded Rules post, and find example disclosures here.

Resources:

Any issues on the community? Report it using the report flag.

Questions? DM the mods!

founded 3 years ago
MODERATORS
ruud@lemmy.world
Loki@lemmy.world
CannaVet@lemmy.world
devve@lemmy.world
ayyy@sh.itjust.works
curbstickle@anarchist.nexus
curbstickle_lw@lemmy.world
32
is K2 horizons historically significant as the first actual open-source LLM, or am I falling for hype? (lemmy.zip)
submitted 16 hours ago by story@lemmy.zip to c/selfhosted@lemmy.world
12 comments fedilink hide all child comments
 

i have been following Chinese models for about two years now because they are open-weight and qwen is fun to run on my kubernetes cluster, the news about this Apache licensed model complete with a recipe to make it again with potentially different ingredients is making we want to abandon them for something more ideologically sound and way more interesting

this new model, couldn't universities rebuild it with different contexts and study it in ways you can't reproduce in other models? like, is k2 horizons a good scientific foundation on which to study machine learning?

or am i just lacking way too much context and falling for hype?

top 12 comments
sorted by: hot top controversial new old
[–] joseki@lemmy.zip 9 points 8 hours ago (1 children)

K2 wasn't the first.

China had CPM-1 in 2020 that was a 2.6B model.

GPT-Neo (Mar 2021) EleutherAI replicates early GPT

PanGu-α (Apr 2021) Huawei's 200B model

WuDao 2.0 (May 2021) BAAI's massive 1.75T sparse MoE model

GPT-J (Jun 2021) EleutherAI release 6B

Meta released OPT in 2022, BLOOM was in 2022, GLM 130B by Tsinghua/Zhipu was in 2022 etc

K2 horizon is very recent, nowhere near the first. Different labs have different computational innovations worth studying. DeepSeek is crazy efficient and doing very novel things. Mistral in France is making very compact local friendly models that are fun and easy to fine tune and merge on consumer hardware.

[–] hendrik@palaver.p3x.de 3 points 1 hour ago (1 children)

Just to add a reference time frame: late 2022 was when ChatGPT got released. I think all these models predate the general population even being aware of AI.

[–] joseki@lemmy.zip 3 points 38 minutes ago

Right and the predecessor to ChatGPT was trained on links from reddit posts pretty wild. I remember having fun making it come up with nonsense prose on my GPU

[–] humanspiral@lemmy.ca 1 points 6 hours ago

This seems to score well on benchmarks. Better than any previous "open blueprint release". While an exact ordered blueprint on static data provides reproducibility, it's not obvious that a different ordering, or more recent versions of webpages/github repos would produce inferior results. I don't really see a huge area of study available from the blueprint, though other labs may find areas of improvement/inspiration in their approaches.

[–] Natanox@discuss.tchncs.de 5 points 9 hours ago (1 children)

My question would be if it uses stuff like CommonCrawl as training material. Given their size I assume they used anything incl. unethical training data, but at least admit it?

The only models I ever found that even tried to only resort to ethically obtained data, being FOSS etc. were the tiny ones from PleIAs. And as expected they're completely useless.

So far I concluded that an "ideologically sound" LLM is impossible due to lack of training data. Unless your ideology allows to steal stuff.

[–] melfie@lemmy.zip 2 points 9 hours ago

I’m assuming a lot of its training data is synthetic and distilled from Chinese models that were themselves trained from pirated data and distilled from American models trained on pirated data. It would be quite remarkable if the training data involved no piracy whatsoever. Then again, it’s open source, so I suppose it would be essentially reversing a reverse Robin Hood.

[–] SuspiciousCarrot78@aussie.zone 26 points 15 hours ago* (last edited 15 hours ago) (1 children)

cries in Swiss

Apertus

But wait...there's even more

[–] hendrik@palaver.p3x.de 10 points 14 hours ago* (last edited 14 hours ago)

Someone ping me once the IFM releases the training data for K2 Horizons and I'll add it to the list. Currently it says "training data will be released to the public" and I can't find any big downloads, so it's not open-source yet.

But sure, big achievement. We only got a handful somewhat recent, usable in size/performance, true open-source LLMs.

[–] PetteriPano@lemmy.world 3 points 11 hours ago (1 children)

I try most models, that get support in llama.cpp or forks. IIRC I wasn't overly impressed with this one. Slower than it should be, mostly because its KV cache is huge.

[–] surewhynotlem@lemmy.world 2 points 8 hours ago (2 children)

I keep falling back to qwen. Though bonsai did give it a good run for a while.

But I only have an old 8gb card.

[–] PetteriPano@lemmy.world 1 points 4 hours ago

I have a bit more memory, so I jump between the two qwen 3.8s.

Glimmer muse was a good contender in the MoE space for a few days. Might be more viable on 8gb.

[–] greybeard@feddit.online 1 points 7 hours ago

Qwen is what I've found to work the best for me so far. I haven't done an exhaustive search or anything, but the mixture of experts 35b models seem to work pretty well.

I've got a little more vRAM to play with, 20GB. It still struggles with giving it enough context to be useful for agentic stuff.