塞巴斯蒂安·拉施卡@rasbt
· 2026-08-11 22:14
塞巴斯蒂安·拉施卡
@rasbt · Sebastian Raschka · 0 followers
Meta 昨天发布了一款全新的开放权重语言模型，这是自过去的 Llama 时代以来未曾发生过的事。他们的 Meta Muse Glimmer 模型是一个……
3今日信号
🔥 Meta Muse Glimmer 发布
1 条最高 7.6
📘 AI Agent 架构
1 条最高 6.6
塞巴斯蒂安·拉施卡@rasbt
· 2026-09-15 21:05
关于基准测试设计的一些思考……这里对 GPT-5.6 Astra 和 Qwen3.8 Max 进行了一次简要的计算机视觉操作（computer-use）对比。任务是……
事实一线研究员正实测对比开源与闭源顶模在 Computer-Use 上的表现。
观点公开基准难以反映复杂 GUI 操作的真实鲁棒性。
对你意味着构建 Agent 必须自建场景实测集，勿盲信榜单。
Some food for thought when designing benchmarks... So, here's a little computer-use (visual) comparison between GPT-5.6 Astra and Qwen3.8 Max. The tas...
📘 大模型训练技巧
1 条最高 4.4
塞巴斯蒂安·拉施卡@rasbt
· 2026-09-11 03:22
很棒的展示，证明在单块 GPU 上也能开展有趣的大语言模型研发工作！
事实Sebastian Raschka 认可单卡完成大模型研发展示的价值。
观点
对你意味着初创团队不必在验证阶段就卷入算力军备竞赛，应推动研发团队用单卡跑通算法与创新 MVP，大幅压低早期试错成本与资本开支。
Nice showcase that interesting LLM work can be done on single GPU!
尚未被主编选入事件/主题的原始推文（共 36 条，按时间倒序）
塞巴斯蒂安·拉施卡 @rasbt09-14 21:24
Since I had to discuss the "pacing" with a lot of people this weekend, here are my two cents: I don't think pacing literally means that these companie...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt08-26 23:05
Now we know: The popular Ox Alpha LLM was GLM-5.3-Flash... Compared to GLM-5.2, this new GLM-5.3-Flash model uses: - a Kimi Linear-style 3:1 (super*) ...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt08-20 09:49
Nice case study on using optimized functions whenever possible (except for educational purposes, though 😆)
🔗 原推塞巴斯蒂安·拉施卡 @rasbt08-15 03:07
RT ACM Education & Learning Center: 8/19, join us for the #ACMTechTalk, "From Conventional LLMs to Reasoning Models to Agents," w/AI & LLM Research En...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt08-06 01:19
RT Hamel Husain: It’s been a long time since I’ve been excited to work through a technical book @rasbt It’s time to bring more ML back into my life
🔗 原推塞巴斯蒂安·拉施卡 @rasbt07-28 21:37
The Kimi K3 architecture figure for yesterday's big open-weight model release, along with some observations and thoughts. 1. Yes, it looks relatively ...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt07-27 23:22
RT elie: this scaling law is a piece of art, kimi K3 recipe improves by ~2.5x over kimi K2 recipe the tech report is amazing
🔗 原推塞巴斯蒂安·拉施卡 @rasbt07-27 21:29
Looking forward to chatting with @hugobowne today (July 27) at 4 pm PT on the Vanishing Gradient livestream on YouTube. Will cover open source, the ne...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt07-26 21:42
Yes, open-source / open-weight models are important for a healthy AI ecosystem. That's how we can verify things, check claims, and keep up outside the...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt07-21 21:11
Thanks everyone for all the kind words and feedback. Super happy that you are enjoying Build a Reasoning Model (From Scratch)! Unfortunately, there's ...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt07-12 00:35
Upon request, here's an updated version with Grok 4.5 and Meta's Muse Spark 1.1. Grok 4.5 seems to sit at the Pareto frontier. Good bang for the buck....
🔗 原推塞巴斯蒂安·拉施卡 @rasbt06-17 08:08
Crazy model! It actually uses the old Qwen2.5-Coder-3B stack and got really great performance with their post-training stack. Need to use it in the ne...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt06-13 20:51
Cool new open-weight model by Cohere: a new lightweight 30B open-weight model for agentic coding tasks. This one builds on Command A+ using the parall...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt06-10 00:31
Always back to the basics: LatentMoE was probably inspired by MLA, which was inspired by LoRA, which was inspired by SVD, which was inspired by eigend...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt06-05 00:41
And another open-weight release. Nemotron 3 Ultra has an ultra impressive capability:efficiency ratio! Design-wise, it carries forward the Mamba-2-att...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt06-04 02:11
It's been a while! 4 nice additions to the open-weight local-LLM-on-consumer-hardware ecosystem:
🔗 原推塞巴斯蒂安·拉施卡 @rasbt05-27 23:07
The MiniMax M2 series was one of the most widely used open-weight LLM series earlier this year. Now, we got a technical report with some interesting t...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt05-23 23:20
Added a DeepSeek Sparse Attention (DSA) from-scratch implementation to my LLMs-from-scratch repo thanks to an awesome new reader contrib. With motivat...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt05-21 07:26
It's been *almost* a bit quiet around LLM architecture releases in the past two weeks 😅 Interesting tidbit is the parallel block design. Via the Cm...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt05-16 21:10
New article: a visual tour of recent LLM architecture advances, from Gemma 4 to DeepSeek V4. I focus on long-context efficiency tweaks like KV sharing...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt05-14 00:38
A little talk on what we can learn from implementing LLM architectures from scratch in Python and PyTorch. And how I approach new open-weight models, ...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt05-13 20:47
Interesting paper. What I like about this is that it is a relatively low-commitment attention modification. I.e., one can use it during most of traini...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt05-10 23:30
Back from a little family break! Lots has happened, and I’m planning to do a deeper dive into the most interesting architectural components (soon). B...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt04-17 03:07
Had a great time at PyCon & PyData DE. Highly recommend it. Great open-source, community-focused conference with lots of builders in the Python AI, LL...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt04-08 22:38
RT Sten Rüdiger: I’ve uploaded a new paper on arXiv (co-authored by @rasbt): MiCA Learns More Knowledge Than LoRA and Full Fine-Tuning In Parameter-...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt04-06 21:07
Added an RSS feed to the LLM Architecture Gallery so it is a bit easier to keep up with new additions over time: https://sebastianraschka.com/llm-arch...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt04-04 21:37
Components of a coding agent: a little write-up on the building blocks behind coding agents, from repo context and tool use to memory and delegation. ...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt03-28 09:54
RT Liquidity Goblin: In an effort to try stop seeing so much slop I've been trying to train my own AI detection model. Found something incredibly inte...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt03-28 06:49
RT levi: Day 83/365 of GPU Programming Looking at DeepSeek's Multi-Head Latent Attention today. The last part of the AMD challenge series is to optimi...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt03-22 20:49
A visual guide to modern LLM attention variants, all in one place: https://magazine.sebastianraschka.com/p/visual-attention-variants
🔗 原推塞巴斯蒂安·拉施卡 @rasbt03-20 01:39
RT Sebastian Raschka: Re The Redbubble poster version just arrived (https://www.redbubble.com/i/poster/LLM-Architecture-Gallery-by-Ahead-of-AI/1792744...
🔗 原推塞巴斯蒂安·拉施卡 @rasbt03-18 10:05
Oh wow, Mamba-3 is here! For me, the most interesting use case of Mamba and Mamba-likes are the recent transformer attention hybrid architectures (Qwe...
🔗 原推
生成于 2026-09-25 22:18（北京时间）· v0.2 CEO 战略情报版