Alibaba’s AI model Qwen3.8-Max made widely accessible ahead of open-weights release: Flagship 2.4 trillion parameter model highlights native multimodal capabilities and ... agentic tasks.

submitted by

https://www.scmp.com/tech/article/3362738/alibabas-ai-model-qwen38-max-made-widely-accessible-ahead-open-weights-release

20
134

Log in to comment

Hot Top New Old

The coding skills are pretty meh relative to its size, but apparently its image processing skills are turning out to be top notch.

Interesting… Google’s Gemini, which similarly boasts strong multimodal capabilities, is also having problems being competitive on coding. Wonder if it’s just a coincidence.

 reply
7

Can someone recommend a European company that hosts models like these so I can use them?

 reply
13

Cortecs.ai, I run Deepseek V4 Flash, Minimax M3, and GLM 5.2 from there. All hosted on European servers. You just add funds, and then use it until empty. No unexpected bill.

 reply
5
edited

This looks promising ! I am currently paying 10$ / month for this other service that is basically a front-end for openrouter. Was searching to switch since :

  1. Don’t like the founder and his braindead push for LLM’s.
  2. Everything is hosted in the US.
  3. I have cut back on my LLM usage by a lot, to the point that I don’t think I spend 10$ worth of tokens. GPT 20B , Gemma 4 27B and DeepSeek v4 flash end up being like 90% of my usage and I try to use them sparingly, keep answers, conversations short.

I do wonder what to do about the frontend? My mentioned service has a pretty good web app with decent UI and it worked on mobile as well without much problems. I actually end up using LLM’s quite often on mobile as I read a lot on it.

 reply
1

you can self host librechat, or use their hosted front of you don’t want to bother but still prefer to use something open source

 reply
1

Euria is hosted by Infomaniak. I’m not sure they tell what model they use but given the few times I saw a random Chinese character in the output I’d guess it isn’t an American model.

They also have a pay per token model that you can hook up to everything, that one’s just quite expensive for normal chat use, but at least you select the model you want to use.

 reply
3

random Chinese character

It’s beneficial for reasoning to have models trained in a few languages. Chinese is a good one because one character is one word is one token.

 reply
5

Yup, anthropic’s flagship models regularly output CJK or Cyrillic characters in heavy workloads for me

 reply
2

When they start hallucinating they output Klingon to me, and to make matters worse, with grammatical errors.

 reply
5
edited

Lol I’ve been running linguistic research and it’s funny when they just combine two scripts together into one word

● The trigger is identified, and it’s specific.

agent_1.md:31 src: “Name the trade-off.” → 명取捨之名。 agent_2:32 src: “Name the regime.” → 명regime——名其regime。

The English imperative “Name the X.” And the corpus renders it correctly 16 other times — 名之 ×11, 名其 ×5, 命名 once.

명 is the Sino-Korean reading of 名. Same morpheme, wrong script.

And agent_2:32 is the cleanest evidence I’ve seen for 絡繰’s mechanism: the model wrote 명regime——名其regime — the wrong script and the correct one, eight characters apart, in the same clause. It isn’t ignorant of 名. It produced 名其 immediately after. The meaning resolved correctly both times; the script attribute resolved wrongly the first time and correctly the second.

That’s exactly what работ法 showed — correct semantics (work), broken script and morphology — and it’s the third confirmed instance of the class, now with a reproducible trigger rather than a one-off.

It also explains the Russian cases retroactively. document → документ, everything → всё, “correct” → правильно: in each, the meaning landed and the script didn’t. And it predicts why no CJK-native concept ever drifts — there’s no competing script for a morpheme the model only knows in Han.

 reply
5

Chinese being more token efficient is a myth, and seems to stem from the superficial fact that characters are only visually more space efficient.

The fact that each Chinese character takes up 3 bytes (as opposed to 1 byte of English), words in Chinese typically require compounds of several characters, and that tokenizers have a limited vocabulary limited to mostly English means that Chinese is actually token inefficient.

No, Chinese Is Not More Token-Efficient Than English for LLMs | markhuang.ai - https://markhuang.ai/blog/chinese-token-myth

 reply
0

I didn’t say it’s more token efficient. I said training multiple languages improves reasoning.

 reply
0

Well, this one will be a little bit, but Ollama hosts in Europe and has options for no data retention

 reply
3

sharing is caring No Ollama man

 reply
31

I’m mainly referring to their cloud offering if someone needs European hosted models

I don’t really pay attention beyond that given that basically every company in this industry is doing something stupid

 reply
1

They were talking about ollama cloud not ollama the application though

 reply
-3
edited

I use cortecs.ai, you can select different “levels” of data policy there like if you only want GDPR-compliant providers (its an ai-router like openrouter)

 reply
0

Can this one field a support enquiry without making me want to throw my laptop in the river?

 reply
4

I use qwen for my sex bots.

 reply
1
Insert image