Damn, Qwen 4 already? I’m still amazed at how good 3.8 27B is and am patiently waiting for 3.8 35B-A3B.
Last time I said I am waiting for the 35b a3b variant apparently they said they were not doing that, they were doing something "better" whatever that means.
I am also eager to see how they react to prism-ml's compression because maybe that way you could trim the size of 27b down to reasonable for us anyway, then 35b a3b becomes obsolete. But we'll have to see, all pretty fresh still.
Qwen3.8 Flash Next is the MoE/new architecture they released instead of a qwen3.8 35b MoE. They didn’t say it’s “better” they just mysteriously said 35B was not what they were releasing this round. A few days later, they released the flash next model weights.
EDIT: and it’s great, doesn’t overthink, but slow AF on my Poweredge T640 server with PCIe3 serving my RTX3090 and DDR4 (even in a proper quad channel config). Best I can get is 15 T/s
legends!
"Wu said Alibaba's Qwen team plans to train a new model with 5 trillion to 10 trillion parameters , to tackle "more complex, longer-horizon tasks" as it pursues an artificial superintelligence, or one that surpasses human capabilities."
"what a time to be alive" -two minute papers
Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.
Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.
As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.
Rules:
Rule 1 - No harassment or personal character attacks of community members. I.E no namecalling, no generalizing entire groups of people that make up our community, no baseless personal insults.
Rule 2 - No comparing artificial intelligence/machine learning models to cryptocurrency. I.E no comparing the usefulness of models to that of NFTs, no comparing the resource usage required to train a model is anything close to maintaining a blockchain/ mining for crypto, no implying its just a fad/bubble that will leave people with nothing of value when it burst.
Rule 3 - No comparing artificial intelligence/machine learning to simple text prediction algorithms. I.E statements such as "llms are basically just simple text predictions like what your phone keyboard autocorrect uses, and they're still using the same algorithms since <over 10 years ago>.
Rule 4 - No implying that models are devoid of purpose or potential for enriching peoples lives.
