OpenAI’s latest large-language model, GPT-5, is winning rapid backing from corporate IT buyers despite a rocky public debut. Oracle said it has embedded GPT-5 across its databases and suite of cloud applications, including Fusion, NetSuite and Oracle Health, allowing customers to pair their internal data with the model’s coding and reasoning capabilities. The integration offers three model sizes through Oracle’s cloud and is designed to automate multi-step processes and boost productivity, the company said. The enterprise momentum follows a surge in direct demand: OpenAI executives told reporters the GPT-5 application-programming interface doubled its traffic within 48 hours of launch, and ChatGPT usage has climbed to roughly 700 million people each week. Chief Executive Officer Sam Altman conceded the rollout was “a little more bumpy than we’d hoped,” prompting OpenAI to restore legacy models, add new Auto, Fast and Thinking modes, and retune GPT-5’s responses to sound warmer and more conversational. Even with those changes, analysts and benchmark testers cited by the Financial Times, Techmeme and Wired say GPT-5’s performance gains over rivals such as Anthropic’s Claude 4.1 and Google’s Gemini 2.5 Pro remain incremental, intensifying debate over whether the current scaling approach to large models is nearing its limits. OpenAI is also courting developers by releasing two open-weight variants, gpt-oss-120B and gpt-oss-20B, under an Apache 2.0 licence—the company’s first open distribution since 2019. The move, coupled with big-ticket enterprise deployments like Oracle’s, signals the firm’s push to broaden GPT-5 adoption even as it works to address user dissatisfaction with the model’s initial behaviour and feature set.