Artificial Intelligence1 min read

OpenAI shares its models' perplexity, or how well they predict data, and proposes pacing limits on such metrics, citing a possible 2027 "avalanche" of advances (Jakub Pachocki/OpenAI)

By · Published by Everything Blog

In short

OpenAI has released perplexity measurements to show how effectively its systems forecast incoming data. Alongside these technical figures, the organization put forward a framework to implement pacing limits on these predictive metrics. Jakub Pachocki highlighted that an avalanche of advanced systems could materialize as early as 2027 without structured boundaries. Regulating these specific indicators is intended to maintain a controlled pace as underlying model capabilities rapidly evolve.

Key points

  • Technical evaluations like perplexity are being used to assess the current pace of machin…: Technical evaluations like perplexity are being used to assess the current pace of machine learning advances.
  • Establishing formal metric limits could help developers systematically govern the timelin…: Establishing formal metric limits could help developers systematically govern the timeline of future model breakthroughs.
  • Anticipated technological acceleration by 2027 has prompted proactive proposals to manage…: Anticipated technological acceleration by 2027 has prompted proactive proposals to manage the speed of industry progress.

Sources checked by our analysis pipeline: 1

Dan Goodin / Ars Technica: |

Lean community blog: |

Jakub Pachocki / OpenAI: |

Stu Woo / Wall Street Journal: |

New York Times: |

Financial Times: |

Jonah Weiner / New York Times: |

Mike Pearl / Gizmodo: |

Financial Times: |

Reuters: |

Tom Wilson / Financial Times: |

Wired: |

Financial Times: |

New York Times: |

This is a Techmeme archive page. It shows how the site appeared at 1:25 PM ET, November 22, 2026.

The most current version of the site as always is available at our home page. To view an earlier snapshot click here and then modify the date indicated.

Artificial Intelligence