Wu Haoning

We are really taking a long time to prove this: everyone is building big macs but we bring you a kiwi🥝 instead. You have multimodal with K2.5 everywhere: chat with visual tools, code with vision, generate aesthetic frontend with visual refs...and most basically, it is a SUPER POWERFUL VLM

多年不见，安兔兔居然也换了一个AI味这么重的前端主页

A basic principle: it is very good to have hypes, but it is essential to have the model truly competitive

@_TobiasLee 已经开始提前帮贵司模型预热了吗

这个 healer alpha 还蛮懂梗的哈哈哈哈

Discover how Kimi K2.5 pushes the boundaries of AI training at scale at #NVIDIAGTC. Join Zhilin Yang, Founder & CEO of @Kimi_Moonshot to explore the strategies driving their massive-scale pre-training. 🗓️ Tues. March 17 | 11:00 a.m. - 11:40 a.m. PDT Add to your schedule: nvda.ws/3OY4bPa

🚨 The @a16z consumer AI Top 100 is back! For the sixth time, we ranked consumer AI websites and mobile apps by usage (monthly unique visits and MAUs). This edition, we changed the rules. Here's why - and what the new list says about where consumer AI is heading 👇

next gen of multimodality belike?

International models on ARC-AGI-2 Semi Private - Kimi K2.5 (@Kimi_Moonshot): 12%, $0.28 - Minimax M2.5 (@MiniMax_AI): 5%, $0.17 - GLM-5 (@Zai_org): 5%, $0.27 - Deepseek V3.2 (@deepseek_ai): 4%, $0.12 These models score below July 2025 frontier labs

@Morty_happy @bigeagle_xd kimi也学习接住了吗

今天最搞笑的事情，昨天初始化龙虾的时候我只接了kimi。 🤔想偷个懒让🦞自己去配置openai。 然后🦞告诉我他已经配置好了，现在都在用openai在聊天。 然后我今天一看日志和账单。 好家伙 kimi伪装成openai一直在给我提供服务。😂

"We develop AI to see a better world not a worse one." Though we as individuals are too trivial to affect that outcome, we can always bear that in mind and try our best to vote for the world we hope to see. We can always choose to do the right thing.

Qwen is totally unrivaled super ultra king for small models.

🚀 Introducing the Qwen 3.5 Medium Model Series Qwen3.5-Flash · Qwen3.5-35B-A3B · Qwen3.5-122B-A10B · Qwen3.5-27B ✨ More intelligence, less compute. • Qwen3.5-35B-A3B now surpasses Qwen3-235B-A22B-2507 and Qwen3-VL-235B-A22B — a reminder that better architecture, data quality, and RL can move intelligence forward, not just bigger parameter counts. • Qwen3.5-122B-A10B and 27B continue narrowing the gap between medium-sized and frontier models — especially in more complex agent scenarios. • Qwen3.5-Flash is the hosted production version aligned with 35B-A3B, featuring: – 1M context length by default – Official built-in tools 🔗 Hugging Face: huggingface.co/collections/Qw… 🔗 ModelScope: modelscope.cn/collections/Qw… 🔗 Qwen3.5-Flash API: modelstudio.console.alibabacloud.com/ap-southeast-1… Try in Qwen Chat 👇 Flash: chat.qwen.ai/?models=qwen3.… 27B: chat.qwen.ai/?models=qwen3.… 35B-A3B: chat.qwen.ai/?models=qwen3.… 122B-A10B: chat.qwen.ai/?models=qwen3.… Would love to hear what you build with it.

Vision results are in for Qwen3.5-397B-A17B, tied for top 2 open model in Vision Arena with Kimi-K2.5-Instant. Ranks #13 overall on par with proprietary models like GPT-4o. Congrats to the @Alibaba_Qwen team!

🧐

希望新的一年可以看到一个更好的世界。 希望新的一年所做的一切能帮助一个更好的世界。 希望我们在喧嚣中能看到文明的进步而非退步。 希望更多的人能幸福。 Happy Year of the Horse.

But we later find out that I/P-frames do not actually split a video accurately; due to their mechanism, they are good for compression but not as near to providing really ‘key’ frames in a video. (2/n)

Interesting take: when I designed the prototype of MoonViT-3D, I was also thinking about I-frame and P-frame and the initial prototype version is based on video chunks divided by I-frames. (1/n)

PDFs are natural ultra-long multi-modal interleaved contents, and they are so important as we use them every day. Multimodal for productivity!

Kimi K2.5 vision support is live and we've updated the Tinker Cookbook to match. Enjoy!

longrun RL 之后聪明的模型难免学会 code switch 的 tendency

@Brian_Bo_Li byok吗 都支持啥品牌的api

我去，太猛了！ 天下苦 Overleaf 久矣，赶 paper 在 chatgpt 和编辑器之间反复横跳真的会疯。prism 出来后又感觉太繁重且不支持中文！ 痛定思痛，我们直接手搓了一个，不管有没有人用，我自己一定会用的产品！ 一个内置 AI Agent 的终极 LaTeX 本地编辑器 LMMs-Lab Writer。 大概有下面几个功能 ✎ 直接内置 AI Agent：不用再复制粘贴，AI 直接在编辑器里帮你写 LaTeX！ ✎ 一键环境配置，自动下载管理 LaTeX，开箱即用。 ✎ 完美解决中文乱码：Ctex？字体？不用管！Agent 自动搞定，妈妈再也不用担心编译出来是空白了。 ✎ 原生 Git 集成：侧边栏直接可视化管理版本控制。 ✎ 本地优先且完全免费：不用担心隐私泄露，没有付费墙，全功能 MIT 协议开源！ ✎ 目前已经支持 Windows / macOS，完全开箱即用。 唯一的下载要求就是去 GitHub 给我们点个 Star 🥺 传送门放评论区了，祝大家论文都 Accept！🚀