GPT-6 Astra feels less lively than 5.6 Sol in everyday chat, user says
The author calls Astra undeniably powerful but says it feels less lively, fun, and human than 5.6 Sol in casual conversation, and worries the new model trades personality for capability.
GPT-6 Astra is undeniably powerful. It has that long missing “big model smell.” 🧠 But when it comes purely to chatting and everyday conversation, compared with 5.6 Sol, Astra definitely feels less lively, less fun, and less human. It reminds me a lot of GPT-5 last year. Compared with 4o, GPT-5 often felt stiff and harder to understand, which triggered a huge wave of negative feedback and eventually pushed OpenAI to bring 4o back. Astra is nowhere near that bad. But I can still clearly feel that the fun, human-like personality OpenAI had finally gotten right with 5.6 Sol has taken a step backward again with Astra. 😿 I know Astra is not primarily a model for emotional conversations. But I still believe a new SOTA model should not improve its ability to get things done at the cost of emotional expressiveness and personality. Those things still matter a lot in human-AI interaction. It’s genuinely disappointing that 5.6 Sol, both Thinking and Instant, was the first generation since 4o that made chatting feel truly fun, natural, and comfortable again. And that personality simply did not carry over to GPT-6 Astra. 😾