Sabbir Hossain

CTO @velox_studio_26 • AI/ML Dev • Trader • Turning ideas into products, GameDev (Hobby)

Building The next gen of education system

[Sabbir Hossain @Sabbirbyte]

@OfficialLoganK Hope so it will be a comeback..

[Sabbir Hossain @Sabbirbyte]

@claudeai Did you guys just copied ot from chatgpt?

[Sabbir Hossain @Sabbirbyte]

@jun_song Mostly people gonna say what they want..but the people who are actually working with the models and in need they are gonna find the right model even if its open source and keep working with it

[Sabbir Hossain @Sabbirbyte]

@bridgemindai 3.5 wasn't keeping up with other models so they skipped it

[Sabbir Hossain @Sabbirbyte]

@0xSero change the config file and your ready to go

[Sabbir Hossain @Sabbirbyte]

@fahdmirza its surely a great thing but the tps is way too low, again cant complain if we are running that kindda model in a consumer hardware

[Sabbir Hossain @Sabbirbyte]

What is graph now!!

[Sabbir Hossain @Sabbirbyte]

@SherryYanJiang Build a dekstop app end to end for model training where you can use cloud system to distribute the model weight in multiple nodes connected through cloud so your friends can join the team and train

[Sabbir Hossain @Sabbirbyte]

@0xSero Which model are you using

[Sabbir Hossain @Sabbirbyte]

@TimJayas They are falling behind without 3.6 flash i guess

[Sabbir Hossain @Sabbirbyte]

@quxiaoyin They will definitely loose the battle then 😂

[Sabbir Hossain @Sabbirbyte]

@dedene Checking its output, thinking through other architecture design

[Sabbir Hossain @Sabbirbyte]

@xmgplus I guess its latest model have already proved what it can do and in future what kind of models with capabilities they can bring..since they dont have enough compute but increasing userbase its the last option they can hop on

[Sabbir Hossain @Sabbirbyte]

@MiaAI_lab Maybe in future versions they will focus on token efficiency rather than a big jump in intelligence like 2.6-2.7

[Sabbir Hossain @Sabbirbyte]

@TrumpTruthOnX Stop poking around and you will be good

[Sabbir Hossain @Sabbirbyte]

@louszbd So the upcoming glm model will fix this?

[Sabbir Hossain @Sabbirbyte]

@ziwenxu_ does it work properly for toolcalls and other usecase? because for me sometime the toolcalls and chat have some parameter/payload issues

[Sabbir Hossain @Sabbirbyte]

🚀 Big news from Alibaba's Qwen team!

Qwen3.8 is launching soon as open-weights — a massive 2.4 trillion parameter model that's already being called one of the most powerful AIs available today. It’s positioned right up there with the frontier leaders, second only to Fable 5.

(Note: the current parameter-size leader on the board is Moonshot’s Kimi K3 at 2.8T — the scale race is heating up fast!)

You don’t have to wait for the full release though! The Qwen3.8-Max-Preview is live right now on Alibaba’s Token Plan, Qoder, and QoderWork. Early access unlocked.

This continues Qwen’s incredible momentum as the world’s most downloaded open-source AI family. Excited to see what the community builds with it once the weights drop.

Who’s jumping in to test the preview? 👀

Links:
International Token Plan → https://t.co/cQO668RlTF
China → https://t.co/E853OeblMs