OpenAI seems to be serious with its internal ‘Code Red’ directive. The latest reports suggest that the AI startup is working on a slew of updates for its chatbot, ChatGPT. As per a Wall Street Journal report citing sources familiar with the matter, the company is said to roll out GPT 5.2 this week, and it will include some big improvements aimed at enterprise users and developers. Further, the report added that the update is being finalised as some staff pushed for a longer timeline to refine the capabilities.
Even after the progress, the code red issued by CEO Sam Altman is not expected to end soon. As per multiple reports, Altman, during interactions with Journalists in New York, confirmed that another major model update is scheduled for January. The upcoming model is said to deliver natural image generation, more expressive personality reminiscent of GPT-4o, and reduced response delays, a combination that may conclude the company’s high-pressure internal push.
Other details have also emerged about the memo that triggered the code red. Altman has reportedly told employees that to improve ChatGPT, they require better use of user preference signals, the same types of feedback patterns previously linked to some sycophantic behaviour in earlier models. He also stated the strategic importance of LMArena, a crowdsourced platform that ranks AI models and has become an informal barometer of public preference. Models that perform on the leaderboard tend to resonate strongly with users, Altman stated.
The intense focus has also exposed a divide inside the company. Product-oriented teams. Led by executives including Applications CEO Fidji Simo and CFO Sarah Friar, the company is said to be pushing for rapid improvements to the everyday ChatGPT experience. On the other hand, researchers focused on artificial general intelligence reportedly favour prioritising breakthroughs in advanced reasoning.
This comes after Google, Anthropic and xAI have introduced their new capable models- Gemini Pro 3, Claude Opus 4.5 and Grok 4.1, respectively.