OpenAI, a leading organization in the field of artificial intelligence, has unveiled its next-generation tools designed to fundamentally transform the developer coding experience. Following the release of the GPT-5 Codex model in September, the company has now announced GPT-5-Codex-Mini, a more compact and cost-effective version. This new move highlights OpenAI's strategy of offering optimized AI solutions for projects of varying scales.
GPT-5-Codex generally provides significant support to developers in tasks such as creating projects from scratch, adding new features to existing codebases, developing comprehensive test scenarios, and performing large-scale code refactoring. In previous tests, GPT-5 Codex was observed to achieve a remarkable success rate of 51.3% in code editing capabilities, outperforming GPT-5 in this area. This model has been integrated into the subscription plans tailored for ChatGPT's enterprise and professional users.
Now, GPT-5-Codex-Mini enters the scene, positioned as an ideal solution, particularly for developers working on smaller-scale projects or aiming to reduce costs. As its name suggests, this model boasts a more compact structure and offers high performance in fundamental coding tasks while also standing out for its cost-effectiveness. OpenAI recommends that developers automatically switch to GPT-5-Codex-Mini when their usage approaches certain limits, thereby ensuring a seamless workflow and budget control.
Various comparative tests have demonstrated the performance of the new model family. According to the results obtained in SWE-bench Verified tests, the GPT-5 High model achieved a success rate of 72.8%, the full version GPT-5-Codex reached 74.5%, and the compact version GPT-5-Codex-Mini secured 71.3%. These figures indicate that while Codex-Mini experiences a slight drop in performance, it stands out for the advantages it offers in terms of cost and ease of use. GPT-5-Codex-Mini is initially available through CLI (Command Line Interface) and IDE (Integrated Development Environment) extensions, with API access expected to be rolled out to users shortly. These developments signify the continually growing role of artificial intelligence in software development processes.