GPT-3 (175B) shows in-context few-shot learning
Confirmed
Importance: historic (5 of 5)The takeaway
OpenAI’s 175-billion-parameter GPT-3 could perform new tasks from a few examples in its prompt, without fine-tuning; it was offered via the OpenAI API from June 2020.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- Historic (5 of 5)
- Last verified
- 29 September 2026
Your AI and this story
- GPT-6 AstraIn its training data
- Claude Opus 5.5In its training data
- Gemini 4 HydrogenIn its training data
- Grok 4.7In its training data
It happened before the cutoffs of all four assistants, so it can be in their training data.
Key facts
- Paper ‘Language Models are Few-Shot Learners’, arXiv 2005.14165 (28 May 2020)
- 175 billion parameters, ~10x larger than any previous dense LM
- Trained on ~300B tokens
- OpenAI API launched in private beta on 11 June 2020
- NeurIPS 2020 best paper award
What happened
GPT-3 demonstrated that scale alone yielded ‘in-context learning’ across translation, QA, arithmetic and writing.
Why it matters
Turned LLMs into a platform; many startups were built on its API, and it directly preceded InstructGPT and ChatGPT.
Sources
2 sources from 2 sites. Numbers match the chips in the text.
2 sources: 2 primary
Primary
- Language Models are Few-Shot Learners (arXiv)arxiv.org, paper
- OpenAI API (OpenAI)openai.com, official
Changes
- Filed