Overview

• OpenAI reinstated GPT-4o as a selectable option for ChatGPT Plus and Pro subscribers after CEO Sam Altman acknowledged that deprecating the model disrupted user workflows.
• The company temporarily doubled GPT-5 message rate limits for paid users to address complaints of sluggish responses and throughput bottlenecks.
• A new toggle labeled “show legacy models” and on-screen indicators now let users choose between models and identify which version generated each reply.
• Engineering teams are actively debugging the internal router that misallocated queries among GPT-5 variants and led to inconsistent output quality.
• Independent red-teaming firms continue to report easy jailbreaks and hallucination vulnerabilities in GPT-5, prompting further safety testing and mitigation efforts.