llms-from-scratch vs hello-agents
Two of the top projects in Learning and cookbooks, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
llms-from-scratch
Code for the book on building a GPT-style LLM in PyTorch
hello-agents
Chinese-language course on building LLM agents from scratch
Verdict
llms-from-scratch ranks higher, 74 to 72. Of the 4 score parts we measure for both, llms-from-scratch leads on 1, hello-agents on 1 and 2 are even.
- Adoptionllms-from-scratch leads by 17 points, 88 to 71: a wide gap.
- FreshnessEven at 100.
- Maintenancehello-agents leads by 12 points, 70 to 58.
- Agent-readyEven at 0.
Pick llms-from-scratch if you are
Python developers learning how LLMs work by coding one in PyTorch
- Main chapter code runs on a conventional laptop and uses a GPU automatically if present
- Plain PyTorch with no external LLM libraries, so every component is readable
Pick hello-agents if you are
Python developers and students learning to build LLM agents, Chinese readers
- 16 chapters from theory to capstone, with runnable code in the code folder
- Includes an Agentic-RL chapter going from SFT to GRPO training
| What we compare | llms-from-scratch | hello-agents |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 88, widely used | 71, popular |
| Freshness | 100, active | 100, active |
| Maintenance | 58, fair | 70, fair |
| Easy to run | Not scored for this kind | Not scored for this kind |
| Agent-ready | 0, none | 0, none |
| Facts from GitHub and the README | ||
| Stars | 106.5k | 82.9k |
| License | custom license (read the license) | custom license (read the license) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | None published | Jul 2026 |
| Language | Jupyter Notebook | Python |
| Docker | No | No |
| GPU | Optional | Not stated |
| arm64 or Apple Silicon | Not stated | Not stated |
llms-from-scratch
Official code for the Manning book "Build a Large Language Model (From Scratch)". Notebooks and scripts implement tokenization, attention, a GPT model, pretraining, classification finetuning, instruction finetuning and LoRA in plain PyTorch, without external LLM libraries. Bonus folders add from-scratch Llama 3.2, Qwen3, Gemma 3, Olmo 3, KV cache, MoE and DPO.
Who it is for: Python developers learning how LLMs work by coding one in PyTorch
Strengths
- Main chapter code runs on a conventional laptop and uses a GPU automatically if present
- Plain PyTorch with no external LLM libraries, so every component is readable
- Bonus material covers KV cache, GQA, MoE, DPO and several model architectures
- Exercise solutions, a companion video course and CI tests on Linux, Windows and macOS
Weaknesses
- Tied to a print book; contributions extending main chapter code are not accepted
- Produces a small educational model, not a production-grade LLM
- License is reported as NOASSERTION; the README states no license terms
- Video course and quiz PDF are hosted by the publisher, not in the repo
- GPU optional
- Needs PyTorch, Jupyter, Ollama (optional, ch07 evaluation)
- Models: GPT-2, Llama 3.2, Qwen3, Gemma 3, Olmo 3
hello-agents
A 16-chapter Chinese tutorial from Datawhale covering agent fundamentals, ReAct, Plan-and-Solve and Reflection, low-code platforms (Coze, Dify, n8n) and frameworks (AutoGen, AgentScope, LangGraph). Later chapters build a custom framework on the OpenAI API and cover memory, RAG, context engineering, MCP/A2A/ANP, Agentic-RL (SFT to GRPO) and evaluation. Chapter code is in the repo's code folder.
Who it is for: Python developers and students learning to build LLM agents, Chinese readers
Strengths
- 16 chapters from theory to capstone, with runnable code in the code folder
- Includes an Agentic-RL chapter going from SFT to GRPO training
- Builds a custom agent framework on the OpenAI API, not only framework usage
- Free online reading and PDF download, plus extra chapters on interviews and skills
Weaknesses
- Written in Chinese; an English README is linked separately
- License is CC BY-NC-SA 4.0, which bars commercial use
- Assumes basic Python and experience calling an LLM API
- Hardware and model requirements for the exercises are unknown
- Needs OpenAI-compatible LLM API
- Models: OpenAI API
More in Learning and cookbooks
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.