Pinned Tweet
Could this be a direction for future model architectures?
Introducing Memory Attention (MA).
alphaxiv.org/abs/2609.28399
Memory Attention is now integrated into Flash Linear Attention (FLA)!
Huge thanks to @uniartisan and @yzhang_cs !
⚡ FLA: github.com/fla-org/flash-lin…
📄 Paper: arxiv.org/abs/2609.28399
J.L retweeted
(1/9)Suppose you're training an LLM on web data with limited math data. Should you place all your math data at the end of training, or mix it with web data earlier? The answer flips with the learning rate (LR) schedule!
At fixed compute and mid-training (MT) token budgets, earlier mixing gives lower MT loss when LR decays throughout training. With constant LR, placeing all MT data for the end works better.
We then vary when LR decay begins. The best tested strategy: keep LR high after MT data enters, then decay near the end.
See our new paper: "Data Schedules and Learning Rate Schedules Should Be Designed Together" !
J.L retweeted
Today we are launching FrontierCS 2, the most frontier benchmark for AI agents in computer science R&D. We built 48 tasks from published papers across 13 areas of CS, refining each one with experts in its field. Excited to co-launch FrontierCS 2 with @kaggle @GoogleDeepMind
J.L retweeted
这个个人网站的设计太帅了！
看了 GitHub 简介，10 月份才开始写的，现在还是那句话：你只需要天马行空，剩下的交给 AI
koimsurai.com
J.L retweeted
你知道的 @teortaxesTex 一直是我们大肥鱼的好兄弟
期待大肥鱼新的表现，具体大肥鱼毕业时间没有人能知道，至少 v41flash 1017 真的很棒，long live RL
我知道听爆料是件很爽的事情但大家有点过度解读了，能说的全是大家知道的消息。
ds4.1pro大肥鱼正在努力学习中具体毕业时间我也不知道。
同时k3.1 pro取名已经明牌flash正在训练中了，很快大家会有便宜的k家模型可以用，可以期待一下。
作为国内我最看好的两家，deepseek和kimi暂时走上了相反的路，其实也代表了deepseek和kimi各自的文化与理念以及困境。
最后说一下我个人看法，国模与O/A两家的差距比我们看到的要大的多，并不是短时间能追上的，‘冲向银河’也只是我对大肥鱼的祝愿以及对ds理念的肯定，同时也希望stepfun，智谱，minimax等众多国模能顶住压力追上去。
J.L retweeted
After much reflection, we’ve redefined FLA’s niche. We want FLA to be the go-to baseline for speed and accuracy—the one everyone wants to compare against—as well as an easy-to-use, reliable default that people feel confident trying.
Alongside K3.1 comes FLA v0.6.0! 💪
Grateful to the open-source community for helping bring faster kernels, more backends, and better KDA.
So proud of my colleagues’ work! ❤️ cc @uniartisan @SonglinYang4
all in RL🫡
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
J.L retweeted
MiniMax M3.1 Flash is out! We trained a DFlash 2 drafter for it, and it's open under Apache 2.0. Generation gets up to 3.5× faster, and it's lossless.
MiniMax M3.1 Flash is out, and we've open-sourced a DFlash 2 drafter for it under Apache 2.0. It makes generation up to 3.5× faster, and it's lossless. Congrats to @MiniMax_AI on the release!
huggingface.co/incoai/MiniMa…
J.L retweeted
Today, we release our first model, Flint-40B-A2B, a fast local coding model built with our automated research system. huggingface.co/RecursiveSI/f…
1. Strong performance. 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2, and 74.2% on SWE-bench Verified.
2. It is trained from scratch (tokenizer->Pretrain->Midtrain->SFT->RL).
3. Fast in your local laptop. Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark.
4. We build the complete model training pipeline in 4 month!
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
J.L retweeted
Hardware looks scary until you break it down into small steps. So I built an app that takes you through Arduino one circuit at a time.
Made with Opus 5.5
learn-arduino-red.vercel.app
你才是真正的openai，友商又要难受了😭
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
