--- title: Build Your Own Reasoning Model: A Full Video Course description: Sebastian Raschka released eight coding lecture videos, totaling 11 hours, that build a reasoning model from scratch in Python and PyTorch. The series starts... canonical: https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Build Your Own Reasoning Model: A Full Video Course | daily.dev og:description: Sebastian Raschka released eight coding lecture videos, totaling 11 hours, that build a reasoning model from scratch in Python and PyTorch. The series starts... og:url: https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm og:image: https://api.daily.dev/og/posts/RugpvvDPM.png og:image:alt: Build Your Own Reasoning Model: A Full Video Course og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Build Your Own Reasoning Model: A Full Video Course **[Sebastian Raschka](https://daily.dev/sources/sebastianraschka)** · 12 min read · 77 upvotes · 2 comments ## Summary Sebastian Raschka released eight coding lecture videos, totaling 11 hours, that build a reasoning model from scratch in Python and PyTorch. The series starts from a pre-trained Qwen3 0.6B model and covers evaluation on MATH-500, inference-time scaling (sampling, self-consistency, self-refinement), reinforcement learning with verifiable rewards using a simplified GRPO implementation, and distillation from a DeepSeek R1 teacher model. The videos accompany his book 'Build a Reasoning Model (From Scratch)' and follow his earlier 'Coding LLMs from the Ground Up' course. He also points to a shorter 90-minute LinkedIn Learning course covering the same concepts at a higher level. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is Group Relative Policy Optimization (GRPO) used for in training reasoning models? GRPO is a reinforcement learning technique used for Reinforcement Learning with Verifiable Rewards (RLVR), where groups of model responses are generated, scored with a verifier, and their relative rewards are used to update model weights. A simplified implementation computes sequence log probabilities and a GRPO loss, then adds clipped policy ratios and KL loss terms to improve training stability. _Readers building RL training loops track implementation deep dives like this one on daily.dev._ ### How can I evaluate a small reasoning model's accuracy on math problems? Build a verifier that extracts the final answer from a model's output, normalizes it, and checks mathematical equivalence against the ground truth, then run it over the MATH-500 evaluation dataset. This same verifier can double as the reward signal for reinforcement learning with verifiable rewards (RLVR) later in training. _Developers comparing model evaluation approaches follow benchmarking write-ups like this on daily.dev._ ### What is the difference between distilling a reasoning model and training it with reinforcement learning? Distillation trains a smaller model on answers generated by a stronger teacher model, such as DeepSeek R1, using supervised training with cross-entropy loss over answer tokens, as an alternative to reinforcement learning. Both soft and hard distillation, and on-policy and off-policy variants, can be used, and results are compared on MATH-500 against reinforcement-learning-trained checkpoints. _Engineers deciding between RL and distillation for reasoning models weigh trade-offs like these on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 60 comments across hackernews (as of 2026-10-26). **TL;DR:** Most commenters praise the course and Raschka's earlier books as excellent, approachable ways to learn how LLMs and reasoning models actually work, while a vocal minority questions the practical value of 'from scratch' tutorials and suspects a personal-branding angle. **Sentiment:** 55% positive · 30% mixed · 15% skeptical **The case for** - Many say the explanations of complex topics like attention, backpropagation, and RLVR are unusually clear and intuitive. - Several report that working through his earlier LLM-from-scratch book made previously confusing concepts finally click. - Some value the hands-on, learn-by-doing approach as a way to truly understand how LLMs work rather than just use them. - A few note post-training or fine-tuning existing open models on custom data is a genuinely practical skill this kind of material supports. **The pushback** - Some argue that producing your own competitive model this way is impractical given the cost and scale requirements, comparing it to a 'build your own CPU' guide. - One commenter frames it as more of a personal-brand monetization play aimed at people who like to feel productive without actually doing anything. - A few feel a gap exists between finishing toy/educational exercises and having any realistic path to apply the skills professionally. **By community** - hackernews (mixed): A heated side-debate about whether 'from scratch' tutorials are genuinely useful sits alongside broad, strongly positive praise for the author's books and teaching style. **Hottest debate:** Whether building a model 'from scratch' via tutorials/videos has real practical value or is mostly a feel-good purchase of identity/personal-brand marketing. **Open questions** - What should someone realistically do with the skills after finishing the book/course — is there a hobbyist path analogous to junior open-source contributions? - What compute resources (local hardware vs. cloud GPUs) are actually needed to work through the material? **Highlights** > I feel there's a misconception in a lot of the comment threads about what an LLM-from-scratch tutorial is for. Of course you're not going to come out of this kind of tutorial being able to train a useful LLM from scratch. (It requires too much compute, and datasets you don't have access to, and there are a lot of tricks and operational problems to solve for real training runs.) If that's your bar for "impracticality", then sure, this is impractical. But that's not the point; the point is understanding these things better by seeing how they are made (or grown, if you prefer). Building something at a level further down the stack than you will ever go forces you to confront details that shape the layers above that you do deal with. This is one reason why Crafting Interpreters is such a beloved book here on HN. LLMs are now at the core of tools we all use everyday... don't you want to know how they are made? Maybe not everyone has this itch to understand the platforms and tools they use, but I sure do, and it's helped me over and over again. Sure, LLMs aren't like operating systems or compilers. They're a quasi-organic mass of billions of parameters learned during training and honestly nobody (not even the mech interp folks) really deeply understands why they work so well. But that's no reason to completely give up and there is still a lot you can learn. Take RLVR ("Reinforcement Learning from Verifiable Rewards"). This is the training technique that made LLMs get so good at programming and math in 2025 and 2026. A lot of people who confidently talk about LLM capabilities either haven't heard of it (and thus end up parroting out-of-date factoids about LLM training derived from the pre-2025 era) or have only a very dim idea of what it is. In this course you actually do RL training of an LLM to get it solving math problems better. You're right down in the details. It's interesting for its own sake but it will also give you a grounding for discussing questions such as: "why did LLM writing get so weird and bad in mid-2026?" and "why is it so hard to train LLMs to be good at things that aren't like math and programming?" Anyway, Sebastian's two books are great and highly recommended; I worked through the first one a year ago and it was one of the best investments of my time in all of 2025. If you prefer video for learning this is great as well (although I'd still say - try the books first). > — [libraryofbabel on hackernews](https://news.ycombinator.com/item?id=50249391) > I'm a bit confused as to who this is for - apart from learning how the tech works that is. If you're truly going to build your own model (and commit to the cost of that), then video tutorials will not cut it. And if you don't really, then I wonder if this is really the best style of conveying the information you'd be interested in? I mean I might be wrong and maybe this helps emerging players or something like that, but to me it is more likely that this is more of a monetisation strategy of a personal brand. Living off selling to people that think they _should_ but then in the end don't. > — [hypfer on hackernews · 6 comments](https://news.ycombinator.com/item?id=50247044) > You are completely off base. His initial book was released almost two years ago. I post train models, for work, and I learned more from his book on LLMs than any other source on LLMs. There is a lot more to it than that book, but you have to start somewhere, even if you have a decent theoretical foundation. There are a bunch of people who learned how to build them from his book. How do you think people learn to build/train models? It comes down from the sky? > — [danielmarkbruce on hackernews](https://news.ycombinator.com/item?id=50248438) > All of these build your own model guides are like someone putting out a build your own CPU guide 10 years ago. Interesting, but not of any practical use. > — [gradus\_ad on hackernews · 2 comments](https://news.ycombinator.com/item?id=50248336) > I'm OK going through the learning itself (will definitely check out the book though, since it's so well liked!). The question is more, what do you do with that once the book or course is done? For a junior coder, there's usually a path to recommend, joining some open source project or doing some home scripting, etc. Something where you can be useful even if it's not your full time job. I'm wondering whether there's some similar hobbyist path available here, or whether there's too much of a jump between the educational/pop sci stuff and the professional realm. If it's the second, that's also fine, I just want to know if I'm missing something. > — [torben-friis on hackernews · 1 comments](https://news.ycombinator.com/item?id=50249162) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50245138) · 409 points · 60 comments ## Community discussion Top comments from developers on daily.dev. **@ristotoldsep** · 6 upvotes > Finally, something deeper than “here is what an LLM is” for the 900th time. nice. **@kartiknvj** · 5 upvotes > Reusing the MATH-500 verifier as the RLVR reward means a normalization bug costs you twice, once as a lower eval score and again as a reward the model learns to game. I'd keep a few dozen hand-checked equivalent answers (fractions vs decimals, units, boxed vs plain) as a unit test on the verifier before kicking off any GRPO run. ## Similar posts on daily.dev - [Build a Reasoning Model From Scratch Is Out](https://daily.dev/posts/build-a-reasoning-model-from-scratch-is-out-vvbpsoxkm) · Sebastian Raschka · 0 upvotes · 0 comments - [AI Reasoning Models Course on LinkedIn Learning](https://daily.dev/posts/ai-reasoning-models-course-on-linkedin-learning-btk2y6txx) · Sebastian Raschka · 0 upvotes · 0 comments - [Do engineers still need to understand how LLMs work?](https://daily.dev/posts/do-engineers-still-need-to-understand-how-llms-work--yqrlcrtga) · Tech World With Milan · 1 upvotes · 0 comments - [Reasoning Models From Scratch: Code Setup](https://daily.dev/posts/reasoning-models-from-scratch-code-setup-n6ro12bbz) · Sebastian Raschka · 2 upvotes · 0 comments --- Tags: [#python](https://daily.dev/tags/python), [#llm](https://daily.dev/tags/llm), [#pytorch](https://daily.dev/tags/pytorch), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Build Your Own Reasoning Model: A Full Video Course","url":"https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm"},"datePublished":"2026-10-24T22:25:24.833Z","dateModified":"2026-10-26T20:35:30.634Z","description":"Sebastian Raschka released eight coding lecture videos, totaling 11 hours, that build a reasoning model from scratch in Python and PyTorch. The series starts...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/97c503e83a714bb328d51a5c51aa760f?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/97c503e83a714bb328d51a5c51aa760f?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Sebastian Raschka","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Sebastian Raschka","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/eb22d2e73d074c5598d31588ad16a51c","url":"https://daily.dev/sources/sebastianraschka"},"commentCount":2,"discussionUrl":"https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":77},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":2}],"keywords":"python,llm,pytorch,reinforcement-learning","timeRequired":"PT12M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Sebastian Raschka","item":"https://daily.dev/sources/sebastianraschka"},{"@type":"ListItem","position":3,"name":"Build Your Own Reasoning Model: A Full Video Course"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm","comment":[{"@type":"Comment","text":"Finally, something deeper than “here is what an LLM is” for the 900th time. nice.","datePublished":"2026-10-26T09:42:08.225Z","url":"https://daily.dev/posts/RugpvvDPM#c-DRTJUeJdO","author":{"@type":"Person","name":"Risto Tõldsep","url":"https://daily.dev/ristotoldsep","image":"https://lh3.googleusercontent.com/a/ACg8ocLDWc6mZn0JwNmXXw6WY0L_HJ6pegzRttooC5VgtXESHSj3MNYx=s96-c"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":6}},{"@type":"Comment","text":"Reusing the MATH-500 verifier as the RLVR reward means a normalization bug costs you twice, once as a lower eval score and again as a reward the model learns to game. I’d keep a few dozen hand-checked equivalent answers (fractions vs decimals, units, boxed vs plain) as a unit test on the verifier before kicking off any GRPO run.","datePublished":"2026-10-26T11:51:26.170Z","url":"https://daily.dev/posts/RugpvvDPM#c-MOSy6oNlH","author":{"@type":"Person","name":"kartik-nvjk","url":"https://daily.dev/kartiknvj","image":"https://media.daily.dev/image/upload/s--3gGgsVCw--/f_auto/v1781456774/avatars/avatar_TvTVeiMdkRCqWUDullFmy?_a=BAMAMiWQ0"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":5}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/build-your-own-reasoning-model-a-full-video-course-rugpvvdpm#faq","mainEntity":[{"@type":"Question","name":"What is Group Relative Policy Optimization (GRPO) used for in training reasoning models?","acceptedAnswer":{"@type":"Answer","text":"GRPO is a reinforcement learning technique used for Reinforcement Learning with Verifiable Rewards (RLVR), where groups of model responses are generated, scored with a verifier, and their relative rewards are used to update model weights. A simplified implementation computes sequence log probabilities and a GRPO loss, then adds clipped policy ratios and KL loss terms to improve training stability. Readers building RL training loops track implementation deep dives like this one on daily.dev."}},{"@type":"Question","name":"How can I evaluate a small reasoning model's accuracy on math problems?","acceptedAnswer":{"@type":"Answer","text":"Build a verifier that extracts the final answer from a model's output, normalizes it, and checks mathematical equivalence against the ground truth, then run it over the MATH-500 evaluation dataset. This same verifier can double as the reward signal for reinforcement learning with verifiable rewards (RLVR) later in training. Developers comparing model evaluation approaches follow benchmarking write-ups like this on daily.dev."}},{"@type":"Question","name":"What is the difference between distilling a reasoning model and training it with reinforcement learning?","acceptedAnswer":{"@type":"Answer","text":"Distillation trains a smaller model on answers generated by a stronger teacher model, such as DeepSeek R1, using supervised training with cross-entropy loss over answer tokens, as an alternative to reinforcement learning. Both soft and hard distillation, and on-policy and off-policy variants, can be used, and results are compared on MATH-500 against reinforcement-learning-trained checkpoints. Engineers deciding between RL and distillation for reasoning models weigh trade-offs like these on daily.dev."}}]} ```