On Monday, OpenAI launches Codexagent coding tools marketed to software developers. Today, OpenAI also released a new model designed for turbo-charging Codex: Codex GPT-5.3.
The company says the model transforms Codex from an agent that can only “write and review code” to one that can do “almost everything that developers and professionals do on computers, developing that can make software and how it works.” After testing the new model against several performance benchmarks, OpenAI claims it can create “highly functional complex games and applications from scratch within days.”
OpenAI says that the GPT-5.3 Codex is also 25 percent faster than the previous model (GPT-5.2) and is the company’s first model that is “instrumental for self-creation,” meaning that company staff use early versions of the program to debug themselves and evaluate its performance.
The release of the new model follows hot on the heels of a new agenic coding model released by its competitor, Anthropic. Indeed, OpenAI and Anthropic originally planned to release the two agentic coding tools at the same time: 10 am PST. However, shortly before the original release time, Anthropic moved the release date up to 15 minutes, slightly ahead of OpenAI in the race to announce the model.