Hacker News front pagePawel Jozefiak4 min readintermediate
Show HN: How Stale Is Your AI? Release age and training cutoff for 20 models
Summary
The author curates release dates and training‑cutoff dates for 20 recent LLMs, exposing how many months each model lags behind its launch. He also provides JSON/txt files and instructions for agents to fetch this data to avoid stale assumptions.
- A table of release and training‑cutoff dates for 20 LLMs (8 labs) is available as HTML, JSON, and plain‑text files.
- Only 10 of the 20 models have a publicly disclosed cutoff, highlighting inconsistent transparency across labs.
- The gap between release and cutoff quantifies staleness; e.g., GPT‑6 Astra released Sep 3 2026 but stopped reading Apr 30 2026 (≈4 months lag).
- Agents can be instructed to fetch https://stale.jock.pl/models.json before naming models, ensuring they don’t rely on outdated knowledge.
LLM‑powered agents need accurate model‑age information to avoid hallucinating current capabilities, and this page supplies a ready‑to‑use source.
5/10
