GitHub announced that OpenAI’s latest code generation model GPT‑5.3‑Codex has been officially pushed to GitHub Copilot. The company said that in multiple benchmarks used internally, the new model showed significant improvements in inference capabilities and execution of complex, long-running workflows that rely on multiple tools. In actual applications, the overall performance of GPT-5.3-Codex on "agent" coding tasks is improved by about 25% compared to the previous GPT-5.2-Codex, significantly shortening developers' waiting time for such tasks.
Currently, to take advantage of this new model, users need to subscribe to a paid plan such as Copilot Pro, Pro+, Business or Enterprise. Supported environments for selecting GPT‑5.3‑Codex include Visual Studio Code (covering chat, question, code editing, and smart agent modes), GitHub.com web, GitHub mobile clients (iOS and Android), the GitHub CLI command line tool, and the GitHub Copilot Coding Agent. If users do not see this option in the interface yet, it means that this update is still in the phase of being pushed in batches.
For users who configure Copilot uniformly in an enterprise or team environment, GitHub explains that organizational administrators using Copilot Enterprise or Copilot Business plans need to explicitly enable GPT‑5.3‑Codex related policies in Copilot settings to allow team members to use the model. GitHub said the upgrade will bring the greatest benefits to developers responsible for designing complex software architectures and automating refactorings, as the new model is specifically optimized for agent-based workflows. By cutting the time it takes to automate toolchain-dependent tasks by about a quarter, this update is expected to significantly alleviate latency issues that have previously been common in large-scale AI-assisted code generation.
This deployment follows the general availability of Claude Opus 4.6 and the launch of its "Quick Mode" preview version. Claude is popular among the developer community for its accuracy in code generation, and the launch of GPT‑5.3‑Codex is seen as strong competition for Claude in the high-end code assistant market. For GitHub Copilot, continuing to introduce multiple models from multiple vendors is a key strategy to maintain its advantage in the highly competitive field of AI coding assistants.
GitHub also encourages users who have enabled GPT‑5.3‑Codex to go to the relevant sections of the GitHub Community to share their experience and feedback. Official documentation and technical guides are expected to be updated after this release to reflect the usage and best practices of the new model.
learn more:
https://github.blog/changelog/2026-02-09-gpt-5-3-codex-is-now-generally-available-for-github-copilot/