OpenAI delays "adult mode"

OpenAI has postponed the launch of its next-generation reasoning model, internally codenamed 'Strawberry' and colloquially referred to as 'adult mode' by some employees. The project, which aims to significantly advance AI's ability to perform deep research, complex problem-solving, and long-horizon tasks, was highly anticipated for a 2024 release. This delay, reported by Reuters and confirmed by internal sources, represents a strategic setback for OpenAI's roadmap to achieve more autonomous and capable AI agents. The company is now focusing on resolving significant technical hurdles related to the model's reliability and safety before a public debut.

The 'Strawberry' model is designed to perform what OpenAI calls 'deep research,' autonomously navigating the internet and using advanced reasoning to solve multi-step problems that are beyond the capabilities of current models like GPT-4. The delay, spanning several months, suggests the challenges are non-trivial, likely involving the model's ability to plan, verify its work, and operate safely without human oversight. This postponement impacts enterprise clients and developers building on OpenAI's platform who were expecting a major capability jump this year. It also intensifies the competitive pressure as rivals like Anthropic's Claude 3.5 and Google's Gemini continue to advance their own reasoning systems.