jingyaogong/minimind
🚀🚀 「大模型」2小时完全从0训练26M的小参数GPT！🌏 Train a 26M-parameter GPT from scratch in just 2h!
This project helps AI researchers and students understand and build small-scale large language models (LLMs) from the ground up. It provides the tools and code to train a functional language model from scratch, starting with raw text data and producing a trained, lightweight LLM. This is ideal for those learning the inner workings of LLM development without needing massive computing resources.
41,159 stars. Actively maintained with 21 commits in the last 30 days.
Use this if you want to learn the fundamental algorithms and training processes behind large language models by building one yourself on consumer-grade hardware.
Not ideal if you primarily need to fine-tune existing, large, production-ready language models or integrate them into applications without understanding their core mechanics.
Stars
41,159
Forks
4,979
Language
Python
License
Apache-2.0
Category
Last pushed
Feb 06, 2026
Commits (30d)
21
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/jingyaogong/minimind"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Recent Releases
Related models
kyegomez/TeraGPT
Train a production grade GPT in less than 400 lines of code. Better than Karpathy's verison and GIGAGPT
theosorus/GPT2-Hasktorch
GPT2 implementation in Haskell with the Hasktorch library, inspired by Andrej Karpathy's Pytorch...
noah-hein/mazeGPT
AI model for making mazes that extends OpenAIs GPT2 model
RohitPawar001/GPT-2-Implementation
This repository contains the implementation of OpenAI's GPT-2 with LORA, QLORA, RLHF, PPO,GRPO,...
miguelvanegas-c/LLMImplementation
This repository demonstrates how to build a functional AI agent using LangChain in Python,...