Qwen3.8 Max: Alibaba counters Kimi K3 without providing evidence

[Redaktion]

Two days after Kimi K3, Alibaba has hit back with Qwen3.8 Max. The company says its 2.4 trillion parameters put it second only to Claude Fable 5. There is no evidence for that so far: no benchmarks, no model card, no license. The preview can still be tested, at one-tenth of the usual price.

https://www.notebookcheck.net/Qwen3-8-Max-Alibaba-counters-Kimi-K3-without-providing-evidence.1347899.0.html

[related to this news]

And a bit related:
notebookcheck.net/Your-own-ChatGPT-offline-AI-without-the-cloud-on-your-laptop.1341872.0.html

Not that one'd be able to run this or Kimi K3 huge open-weight AI LLM models anytime soon locally (but you can use the API), but why this news still matters:

Kimi K3 just fixed 15 critical security bugs that Codex and Fable refused because of "cyber guardrails". Hugging Face: We had this experience ourselves this week! Very scary to be guardrailed as a defender when you know attackers are likely bypassing

[.wojtek]

What's more - being open means certain companies will be able to run those "locally" (for example within EU) without having to relay on Anthropic or OpenAI... more privacy and less revenue for US companies...