Kimi K3 is a triumph of scaling maximalism. They scaled the hell out of everything.
Kimi K3 is a triumph of scaling maximalism. They scaled the hell out of everything.
X / Twitter 6689 stories
reminder that my method for forecasting is literally just greedy decoding
reminder that my method for forecasting is literally just greedy decodingTeortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞): 83% would be very very good. my ranges are 0.7-0.75…
Damn, Xi Jinping's sleeping pills are something fierce, must be the power of traditional Chinese medicine… according to Fable, Kimi K3 is roughly in …
Damn, Xi Jinping's sleeping pills are something fierce, must be the power of traditional Chinese medicine…according to Fable, Kimi K3 is roughly in the ballpark of…
RT Brian McDonald: Noah Smith has built a paid newsletter partly on the proposition that his analysis is worth more than ordinary internet commentary….
RT Brian McDonaldNoah Smith has built a paid newsletter partly on the proposition that his analysis is worth more than ordinary internet commentary.Which makes it fairly…
Local guys hardest hit These models are built for datacenters You would have better luck distilling from them
Local guys hardest hitThese models are built for datacenters You would have better luck distilling from themEric: KIMI-K3 PRISM-DQ (Dynamic Quant) Recipe Analysis is out.I know…
wrong
wrongCan Vardar: i just realized gpt-5.6 sol is literally all i need. openai could stop shipping new models tomorrow and i'd be completely fine
RT He-Man, Principal MLE: "Let's all post a formal outrage response on the same platform with largely known associated accounts, this will surely read…
RT He-Man, Principal MLE"Let's all post a formal outrage response on the same platform with largely known associated accounts, this will surely read as legitimate grassroots…
If this proves to be true, we are in a much more interesting timeline than currently believed
If this proves to be true, we are in a much more interesting timeline than currently believedLeroy Li: 5 units this year actually represent the second…
RT Susan Zhang: open-source is beautiful, indeed
RT Susan Zhangopen-source is beautiful, indeedKimi.ai: We've open-sourced MoonEP, our high-performance communication library for distributed MoE workloads. Built to make expert-parallel communication more efficient at scale,…
I’m not sure if bullying a GDM policy staff on a paper-thin pretext of “we’re mentally ill/boomers and don’t get memes” is a sound strategy of pr…
I’m not sure if bullying a GDM policy staff on a paper-thin pretext of “we’re mentally ill/boomers and don’t get memes” is a sound strategy of…
This will also mint a good deal of new billionaires But we must never forget what happened to Jack Ma…
This will also mint a good deal of new billionaires But we must never forget what happened to Jack Ma…杰克船长宏观策略: 长鑫只是第一张牌。往下翻——长江存储、宇树科技、燧原科技、上海微电子、蓝箭航天、强脑科技。不是一家公司上市，是一个时代的资产排队走上牌桌。长鑫敲锣是DRAM。长江在后面排队——NAND闪存，全球第五，232层量产。宇树——中国版波士顿动力，人形机器人已经进工厂干活了，即将登陆科创板。
It is a hard and sad decision. I shared this message with folks at Thinky. Thank you all for the time together♥️ Just as the last sentence in my mes…
It is a hard and sad decision. I shared this message with folks at Thinky. Thank you all for the time together♥️ Just as the last…
Given how Chinese provincial budgets are battered by the collapse in real estate and LGFV revenues, this will serve as a very, very strong signal for …
Given how Chinese provincial budgets are battered by the collapse in real estate and LGFV revenues, this will serve as a very, very strong signal for…
Two more weeks By H1 2027, China may get *three* successful launch vehicle recoveries (if both ZQ-3 Y2 and CZ-10A work out), from as many separate gro…
Two more weeksBy H1 2027, China may get *three* successful launch vehicle recoveries (if both ZQ-3 Y2 and CZ-10A work out), from as many separate groups.Ace…
We are releasing PerceptionBench, a benchmark that isolates visual perception and evaluates it as a set of atomic capabilities – discovered from how t…
We are releasing PerceptionBench, a benchmark that isolates visual perception and evaluates it as a set of atomic capabilities - discovered from how today's models fail,…
The situation with Fable and Mythos is obscure. I have decently trustworthy info that Anthropic has a larger internal teacher model. But also that Myt…
The situation with Fable and Mythos is obscure. I have decently trustworthy info that Anthropic has a larger internal teacher model. But also that Mythos-Preview was…
Kimi K3 is now available directly from its Hugging Face model page through Together AI, give it a try!
Kimi K3 is now available directly from its Hugging Face model page through Together AI, give it a try!merve: Kimi K3 in @huggingface Inference Providers is…
China is obsessed with logistics, which is a negative izzat field but nevertheless very important. Some Japanese musings in this
China is obsessed with logistics, which is a negative izzat field but nevertheless very important. Some Japanese musings in thisArmchair Analyst: >中国が弱いのはまさに人文社会系Nature Index 2026において、Soicial Science分野の中国のスコアは1,177点で、日本54点の22倍である。自分の思い込みだけで話してしまうから、日本のインテリはレベルが低いのだ。自己評価が高すぎるのだ。若いころにこの人の本を読んでいただけに残念だ。
Making talks with AI agents is awesome. I just told Fable to make a slide with real data on the KL distance from one of our reference Olmo 2 models an…
Making talks with AI agents is awesome. I just told Fable to make a slide with real data on the KL distance from one of our…
RT Vipul Ved Prakash: In partnership with @Kimi_Moonshot, we now have K3 on Together APIs with capacity for hundreds of millions of TPM on launch day!
RT Vipul Ved PrakashIn partnership with @Kimi_Moonshot, we now have K3 on Together APIs with capacity for hundreds of millions of TPM on launch day!Together AI:…
Great overview of K3 from the perspective of history of attention primitives. K3 is very far removed from a "Transformer".
Great overview of K3 from the perspective of history of attention primitives. K3 is very far removed from a "Transformer".ali: http://x.com/i/article/2077616768491585536
RT Ryan Leung 梁興盛, CFA: “当然，造出设备、批量交付、稳定量产是三道不同的门槛。吞吐量、套刻精度、可靠性和维护体系，才决定国产设备最终能替代多少 ASML…
RT Ryan Leung 梁興盛, CFA“当然，造出设备、批量交付、稳定量产是三道不同的门槛。吞吐量、套刻精度、可靠性和维护体系，才决定国产设备最终能替代多少 ASML 订单。”Alpha哥已告诉你什么才是最重要的指标。什么5台/20台不太重要Alpha哥AI投资日记 🌴 📈: 国产DUV光刻机开始量产，半导体设备股集体跳水据 The Information，一家上海国资背景企业已经开始制造国产浸没式 DUV 光刻机，今年计划生产 5 台，2027 年扩大至约 20 台，并准备交付中芯国际、华虹半导体和长鑫存储。若后续能够顺利导入产线，最先受到冲击的将是 ASML 在中国的 DUV 设备业务。
Kimi K3 is now live on @togethercompute! Happy to have Together AI as our day0 launch partner, giving developers immediate access to K3 through high-t…
Kimi K3 is now live on @togethercompute! Happy to have Together AI as our day0 launch partner, giving developers immediate access to K3 through high-throughput inference…
RT Kimi.ai: Kimi K3 is now live on @togethercompute! Happy to have Together AI as our day0 launch partner, giving developers immediate access to K3 th…
RT Kimi.aiKimi K3 is now live on @togethercompute! Happy to have Together AI as our day0 launch partner, giving developers immediate access to K3 through high-throughput…