Models

GPT (Generative Pre-trained Transformer)

OpenAI's family of autoregressive language models, trained on large text corpora then fine-tuned with RLHF. GPT-4o is their flagship multimodal model.

Explained at five levels

Level 1

The name for OpenAI's smart AI that can write, chat, and answer questions — like the brain inside ChatGPT.

Level 2

A series of AI models made by OpenAI that are really good at generating text. GPT-4 is one of the most powerful versions.

Level 3

OpenAI's family of autoregressive language models, trained on large text corpora then fine-tuned with RLHF. GPT-4o is their flagship multimodal model.

Level 4

OpenAI's generative pre-trained transformer series — decoder-only language models trained with unsupervised pre-training followed by supervised fine-tuning and RLHF alignment. The architecture that catalyzed the modern LLM era.

Level 5

A family of decoder-only autoregressive transformers following the scaling law paradigm — exhibiting predictable loss scaling with compute, emergent in-context learning, and post-training alignment via PPO/DPO on human preference data.

Definitions are educational summaries. Terminology can vary by source and context.

Sources