OpenAI Launches Images 1.5, Showcases GPT-5 Wet Lab Optimization

OpenAI announced the launch of Images 1.5, a major upgrade to its image generation capabilities in ChatGPT and the API. The company also showcased GPT-5's ability to optimize molecular cloning protocols, achieving a 79x efficiency gain.

Details

OpenAI kicked off the day with the launch of Images 1.5, a significant upgrade to its image generation capabilities in ChatGPT and the API. CEO Sam Altman touted the new model's ability to produce 'much better images in tons of ways, faster, and new editing capability.' This builds on earlier teasers from Altman and the official OpenAI announcement, which showcased the new 'ChatGPT Images' product experience.\n\nShifting to scientific frontiers, OpenAI researchers unveiled breakthroughs that could accelerate progress in 2026. Specifically, they demonstrated GPT-5's prowess in optimizing wet lab experiments, such as molecular cloning protocols. Partnering with Red Queen Bio, GPT-5 achieved a staggering 79x cloning efficiency gain by iteratively proposing protocol tweaks based on real feedback loops. Remarkably, it even pioneered a novel enzyme-based approach using RecA and gp32 proteins, while piloting autonomous robots for execution under human safety oversight.\n\nAdditionally, a deep-dive thread by Carlos E. Perez challenged the 'stochastic parrot' critique of large language models, showcasing their metalinguistic abilities. Researchers crafted artificial 'toy' languages with novel phonology rules, and while GPT-4, GPT-3.5, and Llama 3.1 struggled, OpenAI's o1 model dominated at 63% accuracy, extracting complex patterns via chain-of-thought reasoning on unseen data.