Astra shows a bigger jump than 5.6 in instruction following and long-chat memory, user says

Qinghua Liu @Qinghua_Liu_Originally posted 2026.09.12

Open original on X

Astra is so damn good. It follows instructions better, remembers long chats, makes great visuals, gets what I mean, catches the little things, and stays organized. 5.6 → Astra is a way bigger jump than 5.4 → 5.5 or 5.5 → 5.6. Older OpenAI models feel like shit now.