September 15, 2025. OpenAI unveiled GPT-5-Codex, a specialized variant of GPT-5 optimized for agentic software engineering inside its Codex product. The company says the model can operate autonomously on large, complex engineering tasks for more than seven hours at a stretch, iterating on implementations, fixing failing tests, and delivering completed work with reduced human intervention.
What GPT-5-Codex does differently
GPT-5-Codex is not simply a faster autocomplete. OpenAI describes it as optimized for “agentic coding”: the model can plan, execute, test, and refine multi-step software tasks inside Codex’s virtual development environment. That includes running tests, diagnosing failures, refactoring, and producing pull-request-style changes that align with typical human review patterns. The company says it trained the model with reinforcement learning on real-world coding tasks to better mirror developer workflows and PR preferences.
