Frontier LLMs still typo: GPT-5.6 outputs "distinguishishable" on high reasoning
stanislavfort · x · 2026-10-12
Researcher stanislavfort spotted a rare typo from a frontier model: GPT-5.6 in high reasoning mode produced "distinguishishable" instead of "distinguishable". He notes it has been a while since he last saw a frontier LLM make such a basic spelling error, making this an odd lapse at high reasoning effort.
More from Fun
- norvid_studies releases simcluster museum tour V on Substack — voooooogel · 2026-10-12
- AI tarot reading demo: 'even the mystics aren't safe' as models deal 3-card spreads — ZeroStateReflex · 2026-10-12
- Dev tells Claude to keep building his game overnight; it replies "sleep well" — prasenx · 2026-10-12
- X users share striking AI-generated cat-eared vampire art in 'QT Your Vampire' meme — Loo_Atreides · 2026-10-12
- Grok hits level 2 in WoW on exploration XP alone — djcows · 2026-10-12
- Grok plays WoW autonomously, now free and heading to Stormwind — djcows · 2026-10-12