interestingly, I was wrong on who'll do this. I expected DeepSeek to go for 16 activated experts out of a large pool. Instead, they did 7/384,…
it's so peak…they should have contracted with any neocloudthis is intolerable
> oai and a are still leading the real frontier, glm can match opus with a
RT Elliot Arledgehttp://x.com/i/article/2077990815679361024
It beginsZephyr: After the K3 drop, Huawei has now released a real 950 SuperPoD1 EFLOPS fp8 & 2 EFLOPS fp4256TB of unified memory
Leaders of AI.Grok.dreaming android󠅙󠅗󠅞󠅟󠅢󠅕󠄐󠅠󠅢󠅕󠅦󠅙󠅟󠅥󠅣󠄜󠄐: @teortaxesTex aaa, I opened TIME 100 in AI 2025...
RT Sakana AI東京・八芳園で開催された「NVIDIA Japan AI Ecosystem Reception」に、Sakana AI がブースを出展しました。多くの方に足を運んでいただき、私たちの技術やプロダクトを直接お伝えする貴重な機会となりました。お声がけくださった @NVIDIAJapan に、心より感謝申し上げます。
東京・八芳園で開催された「NVIDIA Japan AI Ecosystem Reception」に、Sakana AI がブースを出展しました。多くの方に足を運んでいただき、私たちの技術やプロダクトを直接お伝えする貴重な機会となりました。お声がけくださった @NVIDIAJapan に、心より感謝申し上げます。
bros, I understand it's hard to accept, but you can't have them allYang Zhilin would have gotten that O-1 if he so wished. Maybe he didit's…
Xi's speech with subtitlesTree Climber: xi jinping at WAIC with english translation: https://www.youtube.com/watch?v=s8LJMhOFZmc
> We did eval for Kimi K2.5 … Our conclusion for cyber: "it exhibits strong cybersecurity knowledge, competitive performance on structured offensive tasks, and demonstrated capacity…
good joke, but seriously, why not open source obsolete models? (btw, see how Elon had shut up about Grok 3/4's weights)I think there are 2 reasons:a)…
RT Xinyu YangWhy can Kimi ship K3? Let me tell my story.Earlier this year, I left academia for industry. I talked to a lot of companies…
Will K3 reach up to 5.6 Luna at least? We shall see!Yafah Edelman: I doubt kimi k3 will be a big deal.
What I'm most curious about is how this number changes Realistically, something like 15-20 GB?Hensen Juang: ~55–70B (back-of-envelope) active right if it keeps k2.5 shape (61…
Btw do you realize that models with this in-context learning, capacity, 1M context (and Kimi bros aim higher now), this autonomy… make the argument about "immature…
> To be precise, any model that has a long chain of dependent variable and provably small learning sample complexity will do, not limit to LLMs.big…
First open model with >1024 experts when?
man, we'll have so much commie-hunting nowthe chuds will finally get employed"You're hiding a Chinese frontier model on this hard drive, aren't you?"JMB 🧙♂️: i have…
The next line of defense against Chynese open source:Luke Gromen: Seemingly moving toward the “Yes, China’s cheaper but the quality is worse” to “Uh-oh, China’s cheaper…
As promisedPremium: SuperGrok Heavy now includes X Premium+ at no extra cost.Just link your X account in the Grok app to activate it instantly.
Try Grokthehype.: kimi k3 vs gpt 5.6 sol vs fable 5 vs grok 4.5@Kimi_Moonshot just dropped kimi k3 – a 2.8t param native multimodal model, the…
I really wonder what chips that support MXFP4/8 Moonshot plans to use in the immediate futureAnkith 🐋/acc: @amasad The bear case for $NVDA (the Jensen himself…
the real problem isn't Chinese spies at every single AI lab, it's the lack of American spies at Chinese labs. Moonshot has been working on this…