Evolution · 1 min read · Updated 2026-10-05
GPT evolution: from GPT-1 to today
2018 to 2020: bigger and better
GPT-1 (2018) showed that a model trained on lots of text could learn language. GPT-2 (2019) had about 1.5 billion settings and wrote surprisingly smooth text. GPT-3 (2020) had 175 billion, and could do many tasks from just a short instruction.
2022: chat for everyone
ChatGPT launched in November 2022. It was trained to hold a conversation and follow instructions, so anyone could use it with no technical skill.
2023 and after: more skills
GPT-4 (2023) could also understand pictures. Since then, newer models can think through problems step by step, use tools like search, and handle longer documents.
Key takeaways
- Each GPT version was bigger or smarter than the last.
- ChatGPT made the technology easy for everyone to use.
- Newer models focus on reasoning and tools.
Quick questions
What does GPT stand for?
Generative Pre-trained Transformer.
What is a parameter?
One of the adjustable numbers inside the model. More parameters often means more skill.
Do companies still share model sizes?
Often not. Newer models usually do not publish their size.