OpenAI releases agentic coding model amid race with Anthropic

On Monday, OpenAI launched Codex, which is an agentic coding tool marketed specifically to software developers. Today, the company released a new model designed to turbo-charge the capabilities of Codex: GPT-5.3 Codex. According to OpenAI, the new model transforms Codex from an agent that can merely write and review code to one that can perform nearly any task developers and professionals do on a computer. This shift is intended to expand who can build software and how work gets done. Additionally, the company reports that GPT-5.3 Codex is 25% faster than its previous model, GPT-5.2.

The capabilities of the new model extend beyond simple code generation. Having tested its new model against a number of performance benchmarks, OpenAI claims that GPT-5.3 Codex can create highly functional, complex games and applications from scratch over the course of days. Furthermore, according to OpenAI, a company, GPT-5.3 Codex “was instrumental in creating itself.” This means that the company’s staff used early versions of the program to debug the software and evaluate how it was performing during its development.

The launch of GPT-5.3 Codex occurred amid intense, real-time competition with rival AI firm Anthropic, which released its own agentic coding tool. According to the source, OpenAI and Anthropic had originally planned to release their respective agentic coding tools at the exact same time: 10 a.m. Pacific Standard Time (PST). However, not long before the original release time, Anthropic moved its release time up by 15 minutes, slightly besting OpenAI in the race to publicize the models.

Why it matters

The release underscores the intense, real-time competition between OpenAI and Anthropic, as both firms race to dominate the agentic coding market with near-simultaneous product drops.