Discussion

Did Alibaba just kneecap its powerful Qwen AI team? Key figures depart in wake of latest open source release

est: big corp politics.qwen started as the NLP team of Tongyi dept, which was part of the algorithm & model offerings from Alibaba-Cloud (aliyun)Now that's the awkared part: qwen was too successful, a small team had more influence than Tongyi and even Aliyun, and GPUs were scarce, Alibaba had to balance free open source models and customer use.

mellosouls: Subject submitted yesterday fwiw:https://news.ycombinator.com/item?id=47236390

dylan604: timing is everything, but linking back to a submission you made with no comments in a thread with more posts is pointless

carterschonwald: they just released the first small models that i would consider even vaguely articulate for edge inference involving a human. maybe they want to do a mistral and raise a kajillion and work from their home town?

victorbjorklund: What does do a mistral mean?

goldenarm: MistralAI is known for their smaller models on the edge, to avoid competing with Gemini & OpenAI directly.

mycall: Who knows if OpenAI will do a refresh, but gpt-oss-20B/120B are still some of the best edge models so far.

carterschonwald: oh?! what do they handle well? how do they fail?the 3.5 9b model on my laptop at full fp8 is outlandish in its seeming reasoning capacity, though i haven’t really stress tested it

dworks: This is similar to what happened to OnePlus at OPPO. OnePlus was outshining OPPO in the global market - there was absolutely no way the OPPO brand could compete other than price, and its international expansion would have looked like a failure side by side. (Source: Worked there).

bhouston: I feel someone just gave them a huge $$$ offer that they couldn't say no too. Given Elon Musk is praising their efforts, and he lost a lot of his original XAI team recently, my money is on Elon.

teddyX: Nobody with talent wants to work with Elon or xAIThe only people he attracts are h1b candidates who have limited choices

storus: This feels like a typical sociopathic corporate scenario with mushroom management - let a bunch of nerds develop something new/exciting outside mainstream corporate culture, then once it becomes good enough jump in, cut them off and harvest whatever they produced while reaping all benefits/credits for yourself, then live off mediocre subsequent releases for a while while blaming remaining team members for future failures.

cat_plus_plus: Somewhat of a devil's advocate here, because I am very familar with corporate idiocy. But how do you define a non-sociopathic corporate scenario where a company makes a lot of money from a good product they develop? Even if done in maximally practically and emotionally intelligent way, this still requires changes from research phase no?

storus: Developing a business product and monetizing it doesn't require sacking the research group that created it and would have developed it further.

cat_plus_plus: Do you have evidence that they were sacked rather than resigned because they would rather work in a different direction from the one company is taking?

storus: The article directly mentions it; see the section 'Leaving wasn't your choice'.

incomingpain: From what I read, he was fired.Which is insane. Obviously he isnt the only lead at alibaba, but qwen as consequence lost many talented people by doing this. It will negatively impact the Qwen team.Qwen4 is going to flop like Lllama 4.I hope those who just quit form their own new lab and start building again.

NitpickLawyer: > Qwen4 is going to flop like Lllama 4.Maybe, maybe not. It's not even clear if there's gonna be a new qwen, and if they'll keep open sourcing it. It also depends on what the team coming from gemini brings to the table. People move around, and things get shared. Happened before with grok, will likely happen with qwen. Everyone wants what the OG teams have.Mistral was ex llama people. And after their good start, they've kinda plateaued lately. Their latest open models have been quite disappointing. Nothing revolutionary at any rate.People said about the gemini team that moved to xai that they were "amateurs". And yet they delivered in about 1 year with grok4, was SotA for a few weeks at launch. They now lost some people, and likely will get others.Round and round the people move around, and everyone gets most of the things that everyone else uses. I have no doubt that the qwen team will get to find a cozy place to call home for a while...

throwa356262: 'People said about the gemini team that moved to xai that they were "amateurs". And yet they delivered in about 1 year with grok4, was SotA for a few weeks' Wait a minute, this is the same company that is sueing OpenAI for... pretty much this?

vonneumannstan: A bit hard for him to swing if he wants to position xAI as a key Defense contractor for AI and his company is full of Chinese Nationals...

overfeed: Is there a leading American AI research organization - big tech or academia - that isn't "full of Chinese Nationals"? If the DoD want an all-American SoTA model, they may have to wait for a while.