OpenAI introduced the GPT-5.3 Codex model, which is supposed to "upgrade" the Codex agent tool, almost immediately after Anthropic announced its own new product for agent coding. The company promises a noticeable increase in speed and a wider range of tasks that the agent can perform on the computer.
OpenAI introduced the GPT-5.3 Codex model, which is supposed to "upgrade" the Codex agent tool, almost immediately after Anthropic announced its own new product for agent coding. The company promises a noticeable increase in speed and a wider range of tasks that the agent can perform on the computer.
This is reported by TechCrunch. The article says that on Monday, OpenAI launched Codex as an agent tool for developers, and then added a separate GPT-5.3 Codex model to expand the agent's capabilities. OpenAI describes the transition from "write and test code" scenarios to tasks that developers and other specialists usually do at a PC, when not a single answer is required, but a long sequence of actions.
OpenAI also said that GPT-5.3 Codex is 25% faster than its predecessor, GPT-5.2. The company says that it has tested the new model on a number of benchmarks and that the model is capable of creating complex games and applications from scratch within a few days of work. Separately, OpenAI emphasized that this is its first model that was “instrumental in creating itself”: early versions were used to debug the system and evaluate its performance.
The competitive context in this story is also revealing. According to the publication, OpenAI and Anthropic planned to release their agent tools simultaneously at 10:00 PST, but Anthropic moved the announcement up 15 minutes earlier, releasing Claude Opus 4.6 , and formally beat OpenAI to the timing race.
The agent programming market is quickly becoming a direct competitor to large AI labs. Companies are pushing an approach where a model not only generates code snippets but also takes on the multi-step tasks of writing, testing, and debugging to reduce the time from idea to working prototype.
Previously, dev.ua wrote about how OpenAI released a separate Codex application for macOS, which allows you to work with multiple AI agents in parallel: distribute tasks between them, monitor progress, and return to long projects without losing context.
Your Premium trial has ended
