Guide
How long does it take to learn generative AI on YouTube? 19 free courses, measured
Updated 2026-11-19 · 6 min read
A free generative AI course on YouTube takes 2 to 56 hours to watch, and the middle one about 17. We measured the nineteen courses on our list of generative AI and LLM courses on 19 November 2026, counting every video: Krish Naik's MCP sessions are 2 hours 26 minutes, Andrej Karpathy's Neural Networks: Zero to Hero 19 hours 21, and Stanford's CS25 talks 55 hours 37.
What you want decides the hours. Building apps with LangChain, RAG and agents takes 17 to 54 hours of video, depending on the teacher; understanding how large language models work, from a Stanford course or Karpathy's code, 16 to 44. At 1.5× and an hour a day, that is about two to five weeks.
Every course, measured
Ranked by total. "1.5×" is the total watched at one and a half times speed, and "Avg" the average video.
| Course | Videos | Total | 1.5× | Avg |
|---|---|---|---|---|
| Stanford CS25 (talks) | 50 | 56 h | 37 h | 67 m |
| Abu Bakr Soliman: mini-RAG | 26 | 33 h | 22 h | 77 m |
| NPTEL IIT Delhi: LLMs | 38 | 29 h | 19 h | 46 m |
| Stanford CS224N | 23 | 28 h | 18 h | 72 m |
| Stanford CS336 | 18 | 24 h | 16 h | 81 m |
| CampusX: LangGraph | 28 | 24 h | 16 h | 51 m |
| Stanford CS236 | 18 | 24 h | 16 h | 79 m |
| CampusX: LangChain | 21 | 22 h | 15 h | 63 m |
| Karpathy: Zero to Hero | 10 | 19 h | 13 h | 116 m |
| Sheryians: Generative AI | 5 | 17 h | 11 h | 203 m |
| Stanford CME295 | 9 | 16 h | 11 h | 108 m |
| LearningHub: LangChain | 34 | 15 h | 10 h | 27 m |
| IIT Madras: LLMs | 55 | 15 h | 10 h | 16 m |
| Krish Naik: Agentic AI | 14 | 14 h | 9 h | 58 m |
| Kariim Nabiil: n8n | 26 | 11 h | 7 h | 26 m |
| CampusX: Claude Code | 15 | 10 h | 7 h | 39 m |
| Krish Naik: LangGraph | 12 | 9 h | 6 h | 47 m |
| CampusX: RAG | 10 | 9 h | 6 h | 52 m |
| Krish Naik: MCP | 4 | 2 h | 2 h | 37 m |
Totals are rounded to the hour. CampusX's courses are in Hindi, Abu Bakr Soliman's and Kariim Nabiil's in Arabic, the rest in English. All nineteen together are 416 videos and 376 hours. None of the playlists has a private or deleted video. Each course name opens its live page, where the total is current and you can tick off what you have watched. The same courses are on our list of generative AI and LLM courses, with notes on what each covers.
Count hours, not videos
Generative AI courses are made of long videos. Of the 416 here, 168 are an hour or longer, 22 run past two hours and 8 past three. That makes the number of videos a poor guide to the work:
- Sheryians AI School's course is 5 videos and 16 hours 52 minutes: five sessions of 2 to 4 hours each.
- Karpathy's Zero to Hero is 10 videos and 19 hours 21, an average of nearly 2 hours, the longest just over 4. Each one builds a piece of code from nothing, so it is closer to a workshop than a lecture.
- Abu Bakr Soliman's mini-RAG has 6 videos over 2 hours, the longest 4 hours 50.
- IIT Madras is the exception: 55 lectures averaging 16 minutes, 15 hours 2 in all.
The Stanford courses sit in between, at 72 to 108 minutes a lecture: CME295's nine lectures each run 1 hour 41 to 1 hour 51. A two-hour video is a sitting of its own, so split the longest ones over two evenings rather than skipping them.
The machine learning courses that come before these are measured in how long it takes to learn machine learning on YouTube.
How long each route takes
Most people want one of two things: to build apps with language models, or to understand how the models work. These are the totals for common routes made of courses on the list, at 1.5×, watching every day.
| Route | Hours | 1 h a day | 2 h a day |
|---|---|---|---|
| Stanford CME295 | 16 | 11 days | 6 days |
| LearningHub's LangChain and LangGraph, then Krish Naik's MCP | 17 | 12 days | 6 days |
| Karpathy's Zero to Hero | 19 | 13 days | 7 days |
| Krish Naik's LangGraph, Agentic AI and MCP | 25 | 17 days | 9 days |
| IIT Madras's LLMs, then CME295 | 31 | 21 days | 11 days |
| CME295, then CS224N | 44 | 30 days | 15 days |
| Karpathy, then Stanford CS336 | 44 | 30 days | 15 days |
| CampusX's LangChain, RAG and LangGraph | 54 | 37 days | 19 days |
The hours are at normal speed; the days are at 1.5×. Stanford CS25 is left out: it is 50 guest talks to dip into, not a course to finish.
These are days of video. The builder courses are learned by running the code: pause, type it, change it, and fix what breaks, which takes longer than the video. Code on screen is the part not to speed up (what 1.5× and 2× cost →). And these tools move quickly: LangChain and LangGraph courses date within a year, so prefer the newest playlist, while the university courses age more slowly.
Plan the one you choose
Paste the course's link on the home page, or click its name in the table above, and set your daily minutes in the planner: it gives you a finish date and the videos for each day. Tick each video as you finish it and the time left updates. For a route of several courses, add the playlists up to see the whole route at once, and the guide to finishing a long course has the method for the ones with two-hour videos. Every course here assumes Python.
Frequently asked
How long does it take to learn generative AI on YouTube?
The nineteen free courses on our generative AI list take 2 to 56 hours each, measured on 19 November 2026; the middle one is about 17 hours. Learning to build apps with LangChain, RAG and agents takes 17 to 54 hours of video, and understanding how language models work 16 to 44.
How long is Andrej Karpathy's Neural Networks: Zero to Hero?
10 videos and 19 hours 21 minutes, an average of nearly 2 hours a video. At 1.5× it is 12 hours 54, 13 days at an hour a day. Followed by Stanford's CS336, which builds a language model end to end, the route is 43 hours 30.
How long are Stanford's LLM courses on YouTube?
CME295, Transformers and Large Language Models (Autumn 2025), is 9 lectures and 16 hours 10 minutes. CS224N (Spring 2024) is 23 lectures and 27 hours 32, CS336 (Spring 2026) 18 lectures and 24 hours 9, and CS236 on deep generative models (2023) 18 lectures and 23 hours 37.
How long are CampusX's generative AI courses?
Generative AI using LangChain is 21 hours 53 minutes over 21 videos, the RAG playlist 8 hours 39 over 10, and Agentic AI using LangGraph 23 hours 42 over 28: 54 hours 14 for the three in order. Agentic Coding using Claude Code is 9 hours 50 over 15.
What is the shortest generative AI course on YouTube?
Krish Naik's four sessions on MCP, the Model Context Protocol, 2 hours 26 minutes. Then CampusX's RAG playlist, 8 hours 39, and Krish Naik's Agentic AI with LangGraph, 9 hours 21.
Why might a total differ from the playlist today?
These totals were measured on 19 November 2026. Courses add sessions and videos go private. Each course's page here, and the generative AI list, shows the total as it is now.