This page was machine-translated and may differ from the original. View original
OpenAI Unveils Next-Generation Agent-Based Coding Model 'GPT-5.3-Codex'
Significantly enhanced coding and task automation performance, expansion of the integrated Xcode and GitHub development ecosystem
OpenAI has expanded its Xcode and GitHub integrated development ecosystem by releasing GPT-5.3-Codex, the most powerful agent-based coding model released to date.
OpenAI officially launched its next-generation agent-based coding model, 'GPT-5.3-Codex,' on the 6th.
This model combines the frontier-level coding capabilities of the existing GPT-5.2-Codex with the advanced reasoning and expertise capabilities of GPT-5.2, and is reported to be about 25% faster than its predecessor.
OpenAI evaluated this as “the most powerful coding and task automation model released to date.”
GPT-5.3-Codex recorded industry-leading performance across various benchmarks.
In particular, it received top scores on metrics evaluating coding and agent capabilities, such as SWE-Bench Pro and Terminal Bench, and demonstrated strong results in OSWorld and GDPVal, which measure real-world task performance.
OpenAI explained that it verified the actual development and operational capabilities of the model through these four benchmarks.
The interesting point is that the GPT-5.3-Codex is self It is the fact that it is the first model directly utilized for self-development.
The OpenAI Codex team utilized early versions to debug the model training process, automate deployment, and perform testing and evaluation analysis. In other words, the model effectively contributed to its own performance improvement.
The new Codex model is immediately available to paid ChatGPT users on all platforms where the Codex can be used, including apps, CLI, IDE extensions, and web environments.
OpenAI announced plans to provide API access soon. This means developers will be able to integrate the Codex more deeply into their existing workflows.
This announcement aligns with the recently released Codex App.
OpenAI released the Codex app on the 2nd, which allows multiple agents to be managed simultaneously, and the app surpassed 500,000 downloads immediately after its release.
Currently, the weekly active users (WAU) have reached 1 million. OpenAI is providing the Codex to ChatGPT Free and Go plans for a limited time, and has doubled the rate limits for Plus, Pro, Business, Enterprise, and Edu plans.
In addition, GPT-5.3-Codex creates synergy with recently strengthened partnership-based integration.
Through integration with Apple's Xcode 26.3 and GitHub workflows, developers can automate complex tasks by utilizing Codexes directly within the IDE.
OpenAI emphasized, “Codex is evolving beyond a simple code writing and review tool into an agent that handles almost all tasks that experts perform on computers.”
The introduction of the GPT-5.3 Codex is expected to lower the barrier to entry for software creation not only for developers but also for non-experts. At the same time, it is projected to transform development methods themselves and accelerate the expansion of the agent-based development ecosystem.
본 기사에 대한 정정·반론·추후보도 청구는 보도 청구 안내를, 그간 게재된 보도문은 정정·반론보도 모아보기를 참고해 주세요.















