Overview

• Sam Altman conceded the rollout “totally screwed up” and pledged to update GPT-5’s tone to recover the model’s warmth.
• A malfunctioning autoswitcher rerouted many queries to less capable models, leading to inconsistent responses and prompting a routing fix.
• OpenAI reinstated legacy models like GPT-4o for Plus subscribers less than 24 hours after launch to quell consumer anger.
• Enterprise clients such as Cursor, Vercel and Box have already defaulted to GPT-5 for coding and reasoning tasks, praising its speed and lower costs.
• Industry observers suggest GPT-5’s multi-model architecture is aimed at cutting compute expenses, highlighting trade-offs between performance and affordability.