FrontierThe story, in brief

GPT-2: 6-month follow-up

OpenAI releases full GPT-2 (774M params) after 6-month staged rollout—setting new norms for AI safety in publication.

Illustration of a transparent lens revealing connected networks across layers of paper.
Exploring the next frontier of AI research.AI illustration by KeyNews
The KeyNews take

Why it matters

This marks a watershed moment in AI governance: OpenAI's staged release of GPT-2 introduced the 'responsible disclosure' playbook that became industry standard. The decision to withhold and then release the full model shaped how AI labs think about safety vs. openness.

The key facts

9 to know
  1. 774 million parameter model released (full version after 6-month delay)

  2. Staged release: 124M (Feb 2019) → 355M (May 2019) → 774M (Aug 2019)

  3. Open-source legal agreement for model-sharing partnerships included

  4. Coordinated with AI research community on publication norms

  5. Published: August 20, 2019

  6. 774 million parameter model released (full scale)

  7. Staged release over 6 months: 124M (Feb) → 355M (May) → 774M (Aug 2019)

  8. Open-source legal agreement for model-sharing partnerships

  9. Technical report on AI research community coordination and publication norms

Go to the source

OpenAI Blogopenai.com

Publisher excerpt: We’re releasing the 774 million parameter GPT-2 language model after the release of our small 124M model in February, staged release of our medium 355M model in May, and subsequent research with partners and the AI community into the model’s potential for misuse and societal benefit. We’re also…
Read original report
Back to today's editionMore frontier news

Keep reading

Related stories

More from Frontier