OpenAI's API can now keep reasoning across turns instead of discarding it (drop-05a4352b-803.sophisticated-stay.workers.dev)

OpenAI has announced an update to its API, enabling reasoning models like GPT-5.5 and GPT-5.6 to retain reasoning across conversational turns instead of resetting at each interaction. This is a significant advancement for the AI/ML community as it allows models to maintain a more coherent understanding of multi-step tasks, enhancing their effectiveness in complex problem-solving, coding, and scientific reasoning applications. The update introduces the capability for models to continue reasoning from prior interactions, thus improving the continuity and relevance of generated responses. Key features of the update include the introduction of reasoning modes (standard and pro) and a parameter for controlling reasoning effort, which allows developers to specify how much 'thinking' the model should apply based on the complexity of the task. This approach not only optimizes performance and reduces costs for simpler tasks but also enhances the model's ability to perform more intricate workflows. By allowing for retained reasoning via settings like `reasoning.context`, OpenAI enhances the utility of its API in contexts that require in-depth, iterative reasoning, which could lead to broader applications in AI-driven software development and decision-making tasks.