Recorded a podcast, think it’s pretty good and comprehensive, hope you like it ;) youtu.be/EV7WhVT270Q?...

Been a while since I did an LLM architecture post. Just stumbled upon the Arcee AI Trinity Large release + technical report released yesterday and couldn't resist :) Also added a new section to my LLM architecture comparison article with more details: magazine.sebastianraschka.com/i/168650848/20

Been pretty heads-down finishing Chapter 6 on implementing RLVR via GRPO. Just finished, and it might be my favorite chapter so far. Code notebook: github.com/rasbt/reason... (And it should be added to the early access soon.) The next chapter adds stability and performance improvements to GRPO.

For the past month or so, I've been slowly working through this book by @sebastianraschka.com which theoretically and practically builds a GPT model from scratch. Highly recommended! Ironically, I'm writing much more code by hand as a result

Uploaded my State of LLMs 2025 report for this year: magazine.sebastianraschka.com/p/state-of-l... I planned to just write a brief overview, but yeah, it was an eventful year so it was impossible to keep it below 7000 words :D.

One of the underrated papers this year: "Small Batch Size Training for Language Models: When Vanilla SGD Works, and Why Gradient Accumulation Is Wasteful" (arxiv.org/abs/2507.07101) (I can confirm this holds for RLVR, too! I have some experiments to share soon.)

I think of it as this: LLMs lower the barrier of entry, and they make coders (beginners and experts) more productive. It's still worth investing in becoming an expert, because then you will get even more out of LLMs and will be able to deliver even better results.

The LLM eras: 202x Pre-training (foundation) 2022 RLHF + PPO 2023 LoRA SFT 2024 Mid-Training 2025 RLVR + GRPO 2026 Inference-time scaling? 2027 Continual learning?

Just updated the Big LLM Architecture Comparison article... ...it grew quite a bit since the initial version in July 2025, more than doubled! magazine.sebastianraschka.com/p/the-big-ll...

Hold on a sec, Mistral 3 Large uses the DeepSeek V3 architecture, including MLA? Just went through the config files; the only difference I could see is that Mistral 3 Large used 2x fewer experts but made each expert 2x large.

Excited for my first conference in Europe in April. I’ll be talking about LLMs, Python, coding, and all the fun stuff, and I’m looking forward to meeting fellow AI builders there!

This interesting week started with DeepSeek V3.2! I just wrote up a technical tour of the predecessors and components that led up to this: 🔗 magazine.sebastianraschka.com/p/technical-... - Multi-Head Latent Attention - RLVR - Sparse Attention - Self-Verification - GRPO Updates

Looks like we got a new DeepSeek model over the holidays (again): github.com/deepseek-ai/... Basically pushes RLVR & self-refinement to gold-level scores on IMO 2025. Coincidentally, I am currently working on a chapter on self-refinement, and this comes in handy as a nice, scaled-up case study.

Lots of interesting LLM releases last week. My fav was actually Olmo 3 (I love the Olmo series due to their full open-sourceness and transparency). If you are interested in reading through the architecture details, I coded it from scratch here: github.com/rasbt/LLMs-f...

Inference-scaling lets us trade extra compute for better modeling accuracy. Next to RL, it has become one of the most important concepts in today's LLMs, so the book will cover it in two chapters instead of just one. If you are looking for sth to read this weekend Ch4 is available now: mng.bz/Dwra

What should we focus on, (more) LLM training or inference scaling? (A question I got asked multiple times now, so here are some thoughts.) Training is usually very, very expensive, but it is a one-time cost. Inference-scaling is comparatively cheap, but it's a cost we pay at each query.

My "The Building Blocks of Today’s and Tomorrow’s Language Models" talk at the PyTorch Conference is now up on YouTube! youtube.com/watch?v=nDl6... The silver lining of my late arrival and rescheduling: There was no talk after mine, it's followed by a 30 min Q&A instead of just the usual 5 :)

I just saw the Kimi K2 Thinking release! Kimi K2 is based on the DeepSeek V3/R1 architecture, and here's a side-by-side comparison. In short, Kimi K2 is a slightly scaled DeepSeek V3/R1. And the gains are in the data and training recipes. Hopefully, we will see some details on those soon, too.

My new field guide to alternatives to standard LLMs: Gated DeltaNet hybrids (Qwen3-Next, Kimi Linear), text diffusion, code world models, and small reasoning transformers. 🔗 magazine.sebastianraschka.com/p/beyond-sta...

Just saw the benchmarks of the new open-weight MiniMax-M2 LLM, and the performance is too good to ignore :). So, I just amended my "The Big LLM Architecture Comparison" with entry number 13! Link to the full article: magazine.sebastianraschka.com/p/the-big-ll...

A short talk on the main architecture components of LLMs this year + a look beyond the transformer architecture: www.youtube.com/watch?v=lONy...

🔗 Mixture of Experts (MoE): github.com/rasbt/LLMs-f...

Chapter 3, and with it the first 176 pages, is now live! (mng.bz/lZ5B)

Sliding Window Attention 🔗 github.com/rasbt/LLMs-f...

Multi-Head Latent Attention 🔗 github.com/rasbt/LLMs-f...

Just a bit of weekend coding fun: A memory estimator to calculate the savings when using grouped-query attention vs multi-head attention (+ code implementations of course). 🔗 github.com/rasbt/LLMs-f... Will add this for multi-head latent, sliding, and sparse attention as well.

Updated & turned my Big LLM Architecture Comparison article into a video lecture. The 11 LLM archs covered in this video: 1. DeepSeek V3/R1 2. OLMo 2 3. Gemma 3 4. Mistral Small 3.1 5. Llama 4 6. Qwen3 7. SmolLM3 8. Kimi 2 9. GPT-OSS 10. Grok 2.5 11. GLM-4.5/4.6 www.youtube.com/watch?v=rNlU...

From the Hierarchical Reasoning Model (HRM) to a new Tiny Recursive Model (TRM). A few months ago, the HRM made big waves in the AI research community as it showed really good performance on the ARC challenge despite its small 27M size. (That's about 22x smaller than the smallest Qwen3 0.6B model.)

It only took 13 years, but dark mode is finally here sebastianraschka.com/blog/2021/dl...

How do we evaluate LLMs? I wrote up a new article on (1) multiple-choice benchmarks, (2) verifiers, (3) leaderboards, and (4) LLM judges All with from-scratch code examples, of course! sebastianraschka.com/blog/2025/ll...

Just shared a new article on "The State of Reinforcement Learning for LLM Reasoning"! If you are new to reinforcement learning, this article has a generous intro section (PPO, GRPO, etc) Also, I cover 15 recent articles focused on RL & Reasoning. 🔗 magazine.sebastianraschka.com/p/the-state-...

Coded Llama 3.2 model from scratch and shared it on the HF Hub. Why? Because I think 1B & 3B models are great for experimentation, and I wanted to share a clean, readable implementation for learning and research: huggingface.co/rasbt/llama-...

My next tutorial on pretraining an LLM from scratch is now out. It starts with a step-by-step walkthrough of understanding, calculating, and optimizing the loss. After training, we update the text generation function with temperature scaling and top-k sampling: www.youtube.com/watch?v=Zar2...

I'm right now on the last chapter of "Build a Large Language Model (from scratch)" by @sebastianraschka.com and it's absolutely amazing to get started. Now I can understand why people lose their shit over DeepSeek, for example

I just shared a new tutorial: Implementing GPT From Scratch! In this 1:45 h hands-on coding session, I go over implementing the GPT architecture, the foundation of modern LLMs (and I also have bonus material converting it to Llama 3.2): www.youtube.com/watch?v=YSAk...

Yesterday, Google released Gemma 3, their latest open-weight LLM. Finally, a new addition to the "Big 5" of open-weight models (Gemma, Llama, DeepSeek, Qwen, and Mistral). I just went through the Gemma 3 report and experimented a bit with the models, and there are plenty of interesting tidbits:

Just read Gemma 3 is out. Gemma models are super underestimated imho. Will be taking this for a spin in the next few days. In the meantime, they have a technical report here: storage.googleapis.com/deepmind-med...

Just uploaded my "Coding Attention Mechanisms" tutorial. A 2h15m session on coding attention mechanisms to understand how the engine of LLMs works: self-attention → parameterized self-attention → causal self-attention → multi-head self-attention www.youtube.com/watch?v=-Ll8...

I just shared a new article, "The State of Reasoning Models", where I am exploring 12 new research articles on improving the reasoning capabilities of LLMs (all published after the release of DeepSeek R1): magazine.sebastianraschka.com/p/state-of-l... Happy reading!

Takeaways from the latest State of ML Competitions report mlcontests.com/state-of-mac...: - Python & PyTorch still dominate - 80%+ use NVIDIA GPUs, but no multi-node setups 🤔 - LoRA still popular for training efficiency, but full finetuning gains traction. Surprisingly, CNNs still lead in CV comps

@gilesthomas.com is building an LLM from scratch as a Learning exercise and I am so jealous. Working off the book _Build a Large Language Model (from Scratch)_ by Sebastian Raschka First post in series here: www.gilesthomas.com/2024/12/llm-...

A new tutorial in my “Build A Large Language Model From Scratch” series is now live (www.youtube.com/watch?v=341R...) - Tokenizing raw text and converting tokens into token IDs - Applying byte pair encoding - Setting up data loaders in PyTorch for efficient training

Setting up your code environment is often the 1st step when building LLMs for education, research, or production. Made a new video where share my personal setup: "Build an LLM from Scratch 1: Set up your code environment": youtube.com/watch?v=yAcW... (Are videos like this helpful?)

Here’s the 2025 LLM roadmap 😊 1. Code and train your own LLM to really understand the fundamentals 2. Train models more conveniently using production-ready libraries 3. Learn about the big-picture considerations for real-world LLM/AI apps

It's 2025, and I’ve finally updated my Python setup guide to use uv + venv instead of conda + pip! Here's my go-to recommendation for uv + venv in Python projects for faster installs, better dependency management: github.com/rasbt/LLMs-f... (Any additional suggestions?)

Can we merge the query and key weight matrices in an LLM into a single covariance matrix and still train effectively? Here are some promising early results from a reader: github.com/rasbt/LLMs-f... Anyone else familiar with projects that tried this?

Maybe a hot take, but what about the following advice to the next gen: Don't get an AI degree; the curriculum will be outdated before you graduate. Instead, study math, stats, or physics as your foundation, and stay current with AI through code-focused books, blogs, and papers.

quite. btw found @sebastianraschka.com 's simple way refreshing in that regard: "when we say language models 'understand,' we mean that they can process and generate text in ways that appear coherent and contextually relevant, not that they possess human-like consciousness or comprehension"

Just read the s1: Simple Test-Time Scaling paper. Super interesting approach to improving reasoning models! TL;DR: 1. SFT on 1k curated examples w/ reasoning traces. 2. Control response length w/ budget forcing: "Wait" tokens → longer reasoning/self-correction. "Final Answer:" → enforce stopping.

I just finished writing up my take on reasoning models: magazine.sebastianraschka.com/p/understand... Here, I 1. Discuss the advantages & disadvantages of reasoning models 2. Of course, describe and discuss DeepSeek R1 3. Describe the 4 main ways to building & improving reasoning models