OSS
A Show HN tracker puts numbers on model staleness: Gemini 3.1 Pro shipped 13 months after its training cutoff
stale.jock.pl tracks release date against training cutoff for 20 current models across Meta, Anthropic, Mistral, Google DeepMind, OpenAI, Alibaba, xAI and DeepSeek, and hit Show HN at 77 points. The widest gaps are Gemini 3.1 Pro at 13 months (released 2026-02-19, cutoff January 2025), Llama 4 at 8 months (released 2025-04-05, cutoff August 2024) and Grok 4.6 at 6 months (released 2026-08-12, cutoff 2026-02-01). For anyone building on these models, the practical read is that a fresh release number tells you nothing about whether the model knows your framework's current API.
Source
↳ Follow the thread