Sentdex: 'instant' LLM hype rests on pedigree — 150ms is trivial with pure function calling

Sentdex · x · 2026-09-17

Responding to a discussion about low-latency "instant" models, Sentdex argues that much of the trust in them comes purely from the team's pedigree rather than demonstrated value: he hasn't seen a use case where they clearly make sense.

He also points out that "instant" is really 150ms latency, which he says a pure function-calling LLM could match easily — implying the claimed technical advantage is overstated.

Original post →

More from Models

Models channel →