OpenAI introduced Codex on Monday, presenting it as an agentic coding tool aimed at software developers. The very next day, the company unveiled a new model intended to significantly enhance Codex: GPT-5.3 Codex.

According to OpenAI, this model elevates Codex from a tool that simply "writes and reviews code" to one capable of performing "nearly anything developers and professionals do on a computer, expanding who can build software and how work gets done." After evaluating the new model across various performance benchmarks, OpenAI asserts it can develop "highly functional complex games and apps from scratch over the course of days."

The company also reports that GPT-5.3 Codex operates 25 percent faster than its predecessor, GPT-5.2. Notably, this marks the first model that "was instrumental in creating itself," as OpenAI's team utilized early versions of the program to debug its own code and assess its performance.

This new model release comes immediately after a competing agentic coding model was launched by Anthropic. Interestingly, both OpenAI and Anthropic had initially scheduled their respective tool announcements for precisely 10 a.m. PST. However, shortly before that time, Anthropic advanced its release by 15 minutes, narrowly outpacing OpenAI in the public rollout.