Competition in the AI sector intensified this week. After Anthropic introduced Claude Opus 4.6, OpenAI quickly responded by launching its most advanced coding-focused model yet, GPT-5.3 Codex.
According to OpenAI, GPT-5.3 Codex is built to handle more complex and longer-running tasks than earlier models. It goes beyond code generation by reasoning through problems, using tools effectively, and collaborating with users in real time. Anthropic has made similar claims about its latest model, setting up a clear head-to-head rivalry between the two companies.
Designed for Long and Complex Projects
OpenAI says GPT-5.3 Codex combines the strong coding capabilities of earlier Codex models with the advanced reasoning and professional knowledge of the GPT-5 series. This allows the model to manage large projects that can run for hours or even days, while remaining responsive to user input throughout. The company claims it runs about 25 percent faster than its previous version.
Unlike earlier AI tools that delivered results only at the end of a task, GPT-5.3 Codex behaves more like a collaborator. Users can track progress, ask questions, change direction, or request improvements, all while the model maintains full context.
Stronger Performance in Coding and Real-World Use
OpenAI says GPT-5.3 Codex performs well across multiple industry benchmarks focused on real-world software engineering and computer usage. It has shown strong results in terminal operations, file management, operating system navigation, and multi-language programming workflows.
The model has also made notable progress in web development. In internal tests, GPT-5.3 Codex built complete web applications and games from scratch and refined them over time with minimal guidance. Compared with earlier versions, it delivers cleaner layouts, better default settings, and more production-ready features, even with vague instructions.
Meanwhile, Claude Opus 4.6 has achieved the highest score so far on Terminal Bench 2.0, a benchmark that evaluates real-world coding and command-line problem solving. However, OpenAI claims GPT-5.3 Codex has quickly narrowed that gap and offers faster performance and stronger reasoning than GPT-5.2 Codex.
More Than Just a Coding Assistant
OpenAI positions GPT-5.3 Codex as an all-round system rather than a simple coding assistant. The company says the model supports extended research, complex task execution, and advanced tool usage. It can assist across the entire software development lifecycle, including writing documentation, analyzing data, preparing presentations, drafting reports, and supporting research.
In evaluations covering dozens of professional roles, GPT-5.3 Codex matched the performance of OpenAI’s top general-purpose models. As a result, OpenAI believes the model will benefit not only developers, but also designers, product managers, analysts, and researchers.
Security, Availability, and What’s Next
With its increased capabilities, OpenAI has placed greater emphasis on security. GPT-5.3 Codex is the first model the company classifies as high-capability for cybersecurity tasks. It has been trained to help identify software vulnerabilities and includes additional monitoring and access controls to reduce misuse.
GPT-5.3 Codex is currently available on paid ChatGPT plans through the Codex app, command-line tools, and IDE extensions. API access is expected to be released later.
Through this launch, OpenAI signals a shift toward AI systems that do more than write code. The company aims to build tools that can plan, execute, and collaborate across full computer-based workflows. If successful, GPT-5.3 Codex could significantly change how people work with AI, transforming it from a reactive tool into an active collaborator.
That said, it remains unclear which company truly leads the race. Independent benchmark comparisons are still limited, and both OpenAI and Anthropic rely on different testing methods in their public evaluations.