GPT 6 (non Astra) is performing much worse than expected?

A user on Reddit's dev community expressed frustration with the GPT 6 class of models, specifically mentioning that Sol is performing much worse than expected. They noted issues like Sol misinterpreting prompts, rewriting UI elements (e.g., changing graphs into card-based lists), and generally producing poorly thought-out results. The user speculated that 6 Sol might have been renamed to Astra, Terra to Sol, and Luna remained unchanged, which they believe explains the absence of Terra.

This report offers a first-hand account of user frustration with GPT 6 Sol, unlike many reports that focus on its capabilities.