OpenAI announces new model for developers

OpenAI, one of the leading companies in the field of artificial intelligence, has announced a new model. The company, known for developing the ChatGPT chatbot, has introduced the GPT-5-Codex-Mini model. This new model stands out as the latest addition to the Codex series, specifically designed for coding tasks.

Here’s OpenAI’s New “Mini” Coding Model

Last September, OpenAI launched the GPT-5 Codex model. This model was optimized for coding tasks and could operate autonomously. The company stated that it developed this model to improve performance in real-world software engineering tasks. GPT-5 Codex could handle complex tasks such as creating projects from scratch, adding new features to existing projects, and large-scale code editing (refactoring).

GPT-5 Codex significantly outperformed the standard GPT-5 model in code editing tasks. In tests, Codex achieved 51.3% success, while the GPT-5 model achieved 33.9%. This powerful model: ChatGPT Plus was available in paid subscription plans like Pro, Business, Edu, and Enterprise.

The company has now added the GPT-5-Codex-Mini model to this lineup. As its name suggests, this new model is designed with a more compact structure. Developed specifically for small-scale projects and basic tasks, it is said to offer a cost-effective approach.

Performance test results for the new model were also shared. In the SWE-bench Verified test, GPT-5-Codex achieved 74.5% success, while the new “Mini” model achieved 71.3% success. The GPT-5 High model achieved 72.8%. It was also announced that when existing Codex users reach 90% of their limit, the system will automatically recommend switching to the more affordable Mini model.

Developers can currently access GPT-5-Codex-Mini via the CLI (Command Line Interface) and IDE (Development Environment). OpenAI has announced that API support for the model will be available soon. This new, more affordable model is expected to be a valuable option, especially for independent developers and small teams. What are your thoughts on OpenAI’s new coding-focused AI models?

openai