GPT

OpenAI released GPT-1 in 2018, GPT-2 in 2019, GPT-3 in 2020 and GPT-4 in 2023, then moved to several models per generation: larger and smaller variants, multimodal models, and reasoning models that think before answering (the o-series, introduced with o1 in 2024). The GPT models power ChatGPT and are sold through the OpenAI API and Microsoft Azure.

GPT models are decoder-only transformers pre-trained on next-token prediction over large text and multimodal datasets, then post-trained with supervised fine-tuning and reinforcement learning from human and automated feedback. Recent generations combine general chat models and reasoning variants in one lineup, often with several sizes and price points per release. OpenAI also publishes some open-weight models; most GPT models are closed and available only through its API and products.

OpenAI does not disclose parameter counts, architectures or training compute for its recent models, so outside figures are estimates. Reasoning variants are trained with reinforcement learning on tasks with verifiable answers and spend extra inference tokens before the final answer, which are billed as output. Model names and tiers change between generations, so compare models by their measured scores and prices rather than by name.