Ali

@tarat_211 good one

I made a video about my first real attempt at ML research. I started with a simple question about pruning vision-language models, spent weeks breaking Qwen2.5-VL and SmolVLM2 in different ways, hit a bunch of negative results, and eventually submitted my first paper to arXiv.

@foley2k2 @AndrewCurran_ yeah, efficient choices are there

@salysnet @AndrewCurran_ 390b and higher are good for storage. Having a few tens of terabytes of spinning rust works fine since that's faster than your connection. Eventually, ram prices will normalize.

The Trump administration is considering an executive order, and other means, to ban Chinese open-source models within in the United States. Kimi K3 has reignited this debate. Reporting this morning by Axios. Commerce is also considering adding Chinese AI labs to the Entity List.

@RutgerMonty @AndrewCurran_ I wish that it was possible for many of us, but it's for few of us. Maybe some time later people will have

@salysnet @AndrewCurran_ Not for business. The risks will be too high. And if you're talking about end users doing it, well, the consumer hardware to run a 3T model doesn't exist. Maybe when the m8 ultra 3TB comes out.

@Gojjk129942 @FlyaKiet software 2.0

@salysnet @FlyaKiet It’s not exactly the same as software, it includes computation

wdym sold out 😭

@kimmonismus once a model is open-sourced, can it be de-open-sourced? curious to know

The Trump administration is showing signs it could ban cutting-edge Chinese AI models Via Axios The Trump administration is reportedly reviving efforts to restrict cutting-edge Chinese AI models such as Kimi through Entity List designations, procurement rules, security advisories, liability requirements, and public pressure. Measures that could amount to a de facto ban without formally outlawing them.

@AravSrinivas once an open-source alternative becomes "good enough" to handle the vast majority (80%) of standard corporate tasks: the premium justification vanishes, the "good enough" advantages win. the market shifts from prioritizing maximum capability to prioritizing maximum efficiency

@arena @Kimi_Moonshot some real progress. THE AGI IS NEARER.

For the first time, China has taken the lead over the US in Frontend Code Arena with the launch of Kimi-K3 by @Kimi_Moonshot. The last time a Chinese model came close was in early 2025, with DeepSeek-R1.

Open Source AI updates : > Inkling - largest US open-weight model > GLM-5.2 - almost Opus-level > Kimi-K3 - almost Fable-level > Qwen-3.8 - $0.0028/M Expected to beat Opus > Minimax-M3-Pro - 3T, Expected to be Fable-level > much more to come THE AGI IS NEARER.

@tunguz it has a price

Intelligence will be free.

@scaling01 not this time

one last time Messi

Spain!!!

@CommandCodeAI the whole world man

Who wants access to Qwen 3.8 model (2.4T params)?

when a discovery is kept behind a closed door, it is a static asset. When it is open-sourced, it becomes a foundation. we are indeed living through a remarkable acceleration in technology. this rapid evolution you’ve observed-from models with a trillion parameters to those pushing toward four trillion-is a testament to human ingenuity. they are open-sourcing the bicycles of the mind to the whole world.

Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out. Can't wait to hear what you build. Stay tuned! 🚀  Token Plan international：qwencloud.com/pricing/token-… China：platform.qianwenai.com/pricing/token-…

@qwen_cloud it is their screaming era, they will scream alot.

@qwen_cloud interesting time for openai and anthropic and others selling is expensive tokens.

Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out. Can't wait to hear what you build. Stay tuned! Token Plan ： international：qwencloud.com/pricing/token-… China：platform.qianwenai.com/pricing/token-…