OpenAI released a new coding model today, GPT-5.3-Codex. The company said the new model has improved "reasoning and professional knowledge capabilities" and will operate 25 percent faster than its predecessor.
Anthropic, one of OpenAI's chief rivals, also released a new coding model today, Claude Opus 4.6. Earlier this week, OpenAI also launched the Codex app for macOS, a new app interface for managing multiple AI agents at once.
Crucially, OpenAI says that the new GPT-5.3-Codex model is its "first model that was instrumental in creating itself." Anthropic said something similar about Clade Cowork recently, and engineers at OpenAI and Anthropic say almost all their coding is now done by AI.
In a blog post announcing GPT-5.3-Codex, OpenAI wrote, "The Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations — our team was blown away by how much Codex was able to accelerate its own development."
Mashable Light Speed
Why does it matter if an AI model helped create itself?
People who believe in the technological singularity, or "the singularity," talk about a tipping point at which technology becomes self-improving, leading to an uncontrolled explosion of technological advancement. And now we have some real-world examples of AI improving itself. (At least, according to the AI companies behind the announcements.)
We don't know exactly how much GPT-5.3-Codex was involved in its own development, but the news does reveal just how advanced frontier AI models have become, particularly in writing code.
"With GPT‑5.3-Codex, Codex goes from an agent that can write and review code to an agent that can do nearly anything developers and professionals can do on a computer," the OpenAI blog post reads.
GPT-5.3-Codex is available now in the Codex app.
Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.