It may be true for text outputs, but it's absolutely untrue for other modalities: image, speech, voice, and embedding. Heck, embedding models haven't even been updated in years.
OutOfHere · · focus · HN ↗