Models

← All models

InstructGPT 1.3B

OpenAI · 2022-01-27

InstructGPT 1.3B is a 1.3-billion-parameter instruction-following model from OpenAI's January 2022 InstructGPT work. It was trained using supervised fine-tuning on human demonstrations and reinforcement learning from human feedback (RLHF), and demonstrated that smaller RLHF-tuned models can outperform much larger untuned GPT-3 models on instruction-following tasks.

Benchmark scores