On February 6, 2026, OpenAI rolled out GPT-5.3-Codex, which is acclaimed as the most formidable programming agent to date. The launch of this new model happened simultaneously with Anthropic's unveiling of Claude Opus 4.6, sparking an even more intense rivalry in the "AI programming arena" within the enterprise software development sphere. In numerous benchmark tests, GPT-5.3-Codex surpassed Claude Opus 4.6 by a significant margin, scoring 57% on SWE-Bench Pro, 77.3% on TerminalBench 2.0, and 64% on OSWorld. Moreover, it slashed token requirements by over 50% and accelerated reasoning speed by more than 25% for tasks of similar complexity. Transitioning from a mere programming assistant to a proficient operator, this model is now capable of tackling nearly all tasks executed by developers and professionals on computers, thereby aiming for a wider share of the enterprise productivity software market. In addition, GPT-5.3-Codex stands out as the first model to showcase "high proficiency" in cybersecurity-related tasks. OpenAI has implemented a robust protection system and earmarked $10 million in API credits to bolster the development of cybersecurity defense infrastructure.