Recently, OpenAI officially launched its Codex app for the macOS system. This new tool integrates the latest GPT-5.2-Codex model, aiming to improve coding efficiency and compete with the rival Claude Code.

The release of this Codex app comes amid a surge in the development of AI agents. The app allows multiple agents to work in parallel, integrating the capabilities of different agents to adapt to the most cutting-edge workflows. Sam Altman, CEO of OpenAI, stated at the launch: "In complex development scenarios, GPT-5.2 is currently the most powerful model. By encapsulating its strong capabilities into a more flexible interface, we can significantly lower the barrier to entry and bring great convenience to developers."

image.png

Although OpenAI is confident about the performance of GPT-5.2, actual coding benchmark tests have shown different results. Currently, GPT-5.2 ranks first in the TerminalBench test, but the gap with competitors Gemini 3 and Claude Opus is not significant. This indicates that while GPT-5.2 performs well in certain areas, it still needs further improvements in user experience.

The new features of the Codex app provide users with a better experience. It supports automated task setup, allowing users to set scheduled plans to run tasks in the background automatically. The results will be collected in a queue for easy viewing. Additionally, users can choose the interaction method of the agent based on their working style, ranging from rational and practical to empathetic communication, offering flexibility and diversity.