GPT-3 demonstrates broad few-shot learning at 175 billion parameters
OpenAI's original GPT-3 paper reported a 175-billion-parameter autoregressive language model that performed many tasks from instructions or a few examples without gradient updates, while also documenting weaknesses, bias, and human difficulty distinguishing some generated news. OpenAI separately deployed GPT-3-family weights through a controlled private-beta API on June 11.
Why it moved the index
GPT-3 established a major general-purpose capability and scaling milestone, including broad task adaptation and realistic text generation, and its separately dated API deployment verified practical downstream use. Controlled access and low independent agency moderate the assessment.
Assessment history
- R1Toward 47 · confidence 90
New original research milestone with separately verified practical deployment, exact identity, and no durable duplicate.
11 Aug 2026