How should the gap between internal and public AI models be measured?
One commenter estimates a lag of roughly 2–6 weeks for incremental updates and 1–3 months for new pre-trains. Another argues the gap should be measured in capabilities, not months.
TLDR
In a September 26 exchange, one commenter questioned a claim that the capability gap between internal and public proprietary AI models is growing. They estimated an internal-to-public lag of about 2–6 weeks for incremental updates and 1–3 months for new pre-trains. The other commenter argued that, if progress is exponential, a three-month gap in 2026 means something radically different from one in 2024. They also speculated that public models could plateau at Mythos-level capabilities, with all future models kept private because of misuse risks.
How should the gap between internal and public AI models be measured?
One commenter estimates a lag of roughly 2–6 weeks for incremental updates and 1–3 months for new pre-trains. Another argues the gap should be measured in capabilities, not months.
