Back to timeline

GPT-2

OpenAI releases GPT-2, a 1.5B parameter model capable of generating coherent long-form text, initially withheld over misuse concerns.

Model Release

What Happened

OpenAI announced GPT-2, a 1.5-billion parameter language model trained on 40GB of internet text. The model could generate remarkably coherent and contextually appropriate text across a wide range of domains. OpenAI initially withheld the full model due to concerns about potential misuse for generating disinformation, releasing it in stages over the following months.

Why It Matters

GPT-2 was a watershed moment for both AI capabilities and AI safety discourse. It demonstrated that scaling up language models produced qualitatively different behavior — the model could write essays, stories, and even code with minimal prompting. The staged release strategy sparked widespread debate about responsible AI publication practices and set precedents for how the field handles dual-use research.

Technical Details