OpenAI launches GPT-5.3-Codex, its most advanced coding-focused AI model yet
What's the story
OpenAI has unveiled its latest AI model, the GPT-5.3-Codex, a major upgrade in the realm of coding-focused artificial intelligence (AI). The new model is designed to perform complex tasks across the entire software workflow, making it more than just a code generator. The launch comes at a time when competition in developer-facing AI tools is intensifying rapidly. The new model also arrives shortly after its competitor Anthropic unveiled an agentic coding model.
Self-improvement
Model helped in creating itself
What sets the release of GPT-5.3-Codex apart is OpenAI's claim that the model was instrumental in its own creation. This assertion has sparked a lot of interest and debate in the AI community. The company says that early versions of this new model were used by the Codex team to debug training runs, manage deployment, and analyze test results.
Performance boost
Codex now combines coding and reasoning capabilities
The new model combines the coding prowess of its predecessor, GPT-5.2-Codex, with the reasoning and expertise of GPT-5.2. The result is a single model that's approximately 25% faster than before. This performance boost is crucial as Codex is designed to run for long periods, executing tasks involving research, tool usage, debugging, and execution without frequent restarts.
Advanced features
Model can take on long-running jobs
OpenAI describes GPT-5.3-Codex as a "frontier model," capable of taking on long-running jobs that can last over a day. The company also claims the model has improved user intent understanding. For instance, if a vague prompt is given to build a simple website, the model now defaults to adding sensible features and structure, offering a stronger starting point instead of just providing basic results.