LLMilestonesA milestone timeline of AI and large language models

Act I · Foundations

Open modelOpenAI

GPT-2: Too dangerous to release?

More than ten times larger, and able to write coherent long text without fine-tuning.

1.5Bparameters

Why it matters

First showed that scale alone could produce zero-shot abilities, and started the public debate about AI misuse.

Costs and controversy

Released in stages over fears of misuse; the full weights came only in November 2019, which was controversial at the time.

Built on

GPT-1

Led to

GPT-3Mythos Previewgpt-oss

Source: cdn.openai.com ↗

See it on the 3D timeline →