OpenAI has launched Open Responses, a new open-source specification for creating multi-provider, interoperable interfaces for large language models (LLMs). This initiative aims to simplify the integration of different models without needing to overhaul existing systems. Key features include its default support for multiple providers, practical applications for various workflows, and extensibility that avoids fragmentation. Developers can find more details at openresponses.org. The article also highlights the introduction of GPT-5.2-Codex within the Responses API. This model excels at complex tasks such as feature development, code refactoring, and bug detection. It is also noted for its ability to identify vulnerabilities in codebases. OpenAI encourages users to explore their prompting guide for effective usage of Codex models. In addition to these updates, Codex now officially supports "skills," which are collections of instructions and resources that help perform specific tasks. Users can directly invoke these skills or allow Codex to select the appropriate one based on the input. The skills structure follows a simple folder format, making collaboration and sharing easier across tools. Recent enhancements to the Responses API include new connectors that allow users to pull data from services like Gmail and Google Calendar with a single API call. There are also features for maintaining chat histories without managing a separate database, which streamlines the development of interactive applications.