OpenAI just launched GPT-5.3 Codex, a new agentic coding model, only minutes after Anthropic unveiled its own competing tool — turning AI developer tooling into a real-time release race.

The company says GPT-5.3 Codex upgrades Codex from simply writing and reviewing code to acting as a full computer-using agent that can handle “nearly anything developers and professionals do,” including building complex apps and games from scratch over multiple days. OpenAI claims the model is 25% faster than GPT-5.2 and marks the first time one of its models meaningfully helped debug and evaluate itself during development.

The timing wasn’t accidental. Both companies originally planned to release at the exact same moment, before Anthropic nudged its launch 15 minutes earlier — a small move that highlights just how tight and competitive the agentic AI race has become.

The message is clear: agentic coding is the next frontline, and Big AI is no longer taking turns.