OpenAI is preparing to launch its next flagship model, GPT-5, in August, according to reporting and early tester feedback collated by tech news outlet The Information. Testers cited by the publication say GPT-5 markedly outperforms Anthropic’s Claude 4 Sonnet on complex, large-scale software engineering projects and shows higher accuracy in mathematics, physics and other scientific disciplines. One evaluation described the new model as delivering “significant advancements” in practical coding tasks that previously challenged large language models. The model, reportedly codenamed “Zenith” inside OpenAI, is being positioned as a step-change over GPT-4 and recent open-source releases. Its expected debut underscores the intensifying competition among leading AI developers to push generative systems deeper into professional software development and research workloads.