GoKawiilNew tracker catalogs training cutoff and release gap for 20 AI models
A new project called 'How Stale Is Your AI?' compiles release dates and training cutoffs for 20 models from 8 major labs, sorted from stalest to freshest. It finds that only 9 of the 20 models have a training cutoff date that is actually published by their maker, with data also offered as a downloadable models.json file.
GoKawiil's interpretation of the reporting above, not reported fact.
Users often assume a model's knowledge is as current as its release date, but the gap between when training data ends and when a model ships can span months, and web search features don't close that gap since they only retrieve information temporarily rather than updating the model's underlying knowledge. This kind of transparency effort highlights how inconsistent AI labs are about disclosing a basic fact that affects how much users should trust a model's answers about recent events.
- Only 9 of 20 tracked AI models have a training cutoff date officially published by their maker.
- A model's release date can lag its training cutoff by many months, meaning it's already outdated at launch.
- Web search or browsing features don't update a model's core knowledge, they just temporarily supplement single answers.
Source: stale.jock.pl — Pawel Jozefiak, 2026-09-16
Published there as: “Show HN: How Stale Is Your AI? Release age and training cutoff for 20 models”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.