OpenAI Launches GPT-5 Version For Codex Pengoding Agent

JAKARTA OpenAI launched the latest GPT-5 model for Codex, an Artificial Intelligence (AI)-based coding agent. This latest AI model is claimed to have more dynamic thinking skills.

GPT-5-Codex is now available to all paid users on the Codex platform. This includes ChatGPT Plus, Pro, Business, Edu, and Enterprise customers. OpenAI also plans to provide this model for API customers in the future.

This coding-specific AI model is designed to challenge similar products from other companies. Some of the competitors from other AI companies include Claude Code, Cursor from Anysphere, and GitHub Copilot from Microsoft.

The increase in GPT-5-Codex performance has indeed increased significantly. In the coding benchmark, this model managed to outperform the previous version. This model is also better at handling code refacturing tasks from large repositories.

Alexander Embiracos from OpenAI, launching from TechCrunch, stated that this model can adjust the work time of a task in real-time. The time needed can range from a few seconds to seven hours.

This ability is considered superior to traditional router systems. Routers usually determine the cloud power and the time needed at the beginning, but GPT-5-Codex can decide halfway to take more time.

In addition to coding, the model is also trained to conduct code reviews. A number of software engineers have evaluated comments made by this model and the results show that GPT-5-Codex provides fewer wrong comments.

This improvement shows that OpenAI has succeeded in making Codex more effective and reliable. With flexible thinking skills, this model can handle more complicated coding problems.