Ian Li

@craigweiss 哈哈哈，不过中文还真是挺难的

i started learning chinese faster, just in case

@jun_song I feel like part of the reason he constantly sends this is just pride in his own work.

Be Elon Musk: > Become the world’s first trillionaire > Get 230M followers > You still have to post ‘Try Grok’ 10 times every single day

@arkuy99 支持

傻逼公司

@AlphaEmmanuel15 @jun_song 很难很难

@jun_song Are you sure about the numbers? 200 out of 14 million is severely BRUTAL!!! Like they retain the best of the best of the best of the best of the best and on and on. That's the uppermost layer of the upper echelon. Like the VVIP group within a VVIP's VVIP. DAMNIT China. Impressive

Why Tsinghua University keeps dropping AI breakthroughs: Chinese kids take math and CS classes from a very young age, and the college entrance competition is brutal. Right now, Computer Science is the absolute top-tier, most desired major. They only accept about 200 students per year for CS-related majors. The number of competitors? 14 million. That’s a 0.001% acceptance rate. Unless you are a literal genius, getting in is near impossible. Personally, I think the competition there is way more insane than any university in the US.

@thsottiaux I agree with this idea

GPT-5.6 Sol confirmed to be an extremely good model

lol……Alright, 5.6 is a good model.😂

@thsottiaux God

Oops... I did it again. Enjoy reset usage limits for all paid users for Codex and ChatGPT Work. Super grateful for an incredible team who is iterating at lightspeed and keeping the infra up as we scale faster than ever. Enjoy the weekend!

@kunchenguid Thanks for sharing, I think I need to reconsider whether to recharge it.🤔

ok just spent a morning with Kimi K3 as my firstmate, here's my real experience 1. it's very, very slow potentially due to the fixed max reasoning. you should expect the experience of something slightly slower than fable 2. its claimed cost efficiency is not manifesting in real economics i bought the $40 plan, and a few prompts later it's already eaten 1/3 of my 5-hr limit - it was in a single session and my context window was only 200k long at that time i don't care what the benchmark numbers say, and what the face value API pricing is, in reality Kimi K3 burns my Kimi subscription as quickly as Fable burns my Anthropic plan - i observe no efficiency benefit 3. its instruction following capability is weaker than other frontier models firstmate stretches frontier models' reasoning capability and is a really good test that can quickly reveal how good a model is at following instructions the pure "intelligence" of K3 does hold up - it understands my intent very well, and can diagnose problems, delegate tasks all fine but i very quickly noticed many instructions in firstmate's system prompt not strictly followed by Kimi K3. these were never a problem with gpt 5.5, 5.6, opus, fable and grok 4.5 so all in all, i'm now very skeptical of the claimed performance and going to keep my eyes wide open on its true capability

@henghaer123 @grok 国产卡能否训练出最新的 kimik3 模型？是否有这个能力？

我是不太理解，国内到底偷运进来多少英伟达的卡，能让Kimi训练个这么猛的大模型🤦

@MaxForAI 哈哈哈哈哈

中国人太能偷了，他们竟然通过蒸馏偷出了尚未被发布的出来的美国前沿模型。

@petergostev 牛逼！

Let me show you how to run Kimi-K3 locally, right from your own house

@MaxForAI @justinsuntron 我草，666

如果我是你，我就会和 @justinsuntron 一起关注@MaxForAI 这个账号😎

@xicilion 我也是

@justinsuntron 太强了

传说中的Kimi K3上了

@MaxForAI @GoogleDeepMind 查无此人，就连快要发布了感觉都无人在意。

让我提醒你，这张图里没有任何一个Gemini模型 @GoogleDeepMind 你们在做什么？

Kimi K3 is here. A 2.8-trillion-parameter model with native vision, a 1M-token context window, and full open weights coming soon. But this is not impressive simply because of its size. Across coding, agentic search, knowledge work, office tasks, and vision benchmarks, Kimi K3 is now genuinely competitive with the world’s strongest proprietary models—outperforming GPT-5.6 Sol, Claude Opus 4.8, and even Fable 5 on several evaluations. It is not the best model at everything, and Moonshot is refreshingly honest about that. But seeing an open model built in China enter the global frontier is something worth celebrating. Congratulations to the Kimi and Moonshot AI team. Chinese AI is no longer simply catching up. It is increasingly helping define the frontier. 🇨🇳🌙

@ivanfioravanti @MiniMax_AI @Zai_org Have fun.

Planning (at least) three weeks in China in September with my wife and young daughter that speaks Chinese! ✈️ I hope to be able to visit at least @MiniMax_AI and @Zai_org labs while there 🙏

@attaalong @bakkermichiel 正确的，我真没想到的是，国内一群二货也完全相信蒸馏叙事，认为中国纯抄袭，认为国外的都是“自研”🤗

@bakkermichiel 兄弟，你可能不是做这个行业的人。蒸馏只是一种手段，但是你想通过蒸馏就能够去超越比肩最 SOTA 的 model，这是不可能的。如果说这样能做到的话，为什么只有中国能做到？其他国家去哪呢？其他国家既没有 GPU 的限制，也没有访问美国模型的限制。

Can someone explain me how the new Kimi can be so insanely good? The story has been "chinese labs have much less compute but they still stay close to the frontier through distillation". These results seem impossible to explain through distillation alone.

@sama Come on！

we did not have our best last 12 months ever, which is mostly my fault, but we are about to have our best 12 months to date. the team is doing amazing work and i think you’ll be very happy with what they’ve got cooking for you. i am happy about this for many reasons, but mostly because i care about our users winning. AI has to be about giving lots of people more freedom, agency, and wealth. we want to do the right thing, but we do not want to scare people into doing our thing.

@kunchenguid Insane numbers. Looking forward to your hands-on review 🤩 Great to see Chinese open source models keep getting stronger.

kimi k3 looks incredible.. if everything here is true, it’s a model with fable level intelligence at sonnet level pricing but benchmark numbers are not representative enough of how good the model feels. i’ll have to try it out!