AI Hub
All models
LLMOpen weights

GPT-1

OpenAI

Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error

Superseded by GPT-2

Specifications

Type
LLM
Access
Open weights
Released
June 11, 2018
Parameters
117M
Context window
512 tokens
Input
Text
Output
Text

Our take

The first GPT: a Transformer decoder pretrained on unlabeled text then fine-tuned per task. It introduced the generative pretraining recipe that everything since builds on.

Related research

Compare GPT-1 with

See all GPT-1 alternatives →