GPT
OpenAI's line of language models, including GPT-2, GPT-3, GPT-3.5, and GPT-4; the original foundation models that started the LLM era.
Think of it like
The family that changed everything; now synonymous with "large language model."
Example
GPT-3 (2020) had 175B parameters and showed that scale unlocked new abilities; GPT-4 (2023) added multimodal input.
How it actually works
GPT models use a causal transformer trained on next-token prediction, then fine-tuned with RLHF. Each new version has been trained on more data, with bigger models and more compute. GPT-4 is a multimodal gated release.
For product teams
GPT established the API + chatbot product model; pricing is per token (input + output).
For engineers
GPT-4 is frontier: state-of-the-art reasoning, code, multimodal. Inference is expensive; quantization and distillation popular.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome