BroadChain learned that on April 25, OpenAI announced the launch of GPT-5.5, calling it the most intelligent and intuitive model to date, now available on API, Codex, and ChatGPT. This model can understand user intent faster, take on more work tasks, and significantly reduce the number of tokens used to complete the same tasks, improving efficiency.
The core capabilities of GPT-5.5 include:
Agentic Programming: The most powerful agentic programming model to date, with clear advantages in Codex, capable of handling engineering tasks such as implementation, refactoring, debugging, testing, and verification.
Knowledge Work: Combined with Codex's computer use skills, the model can work alongside users to operate computers, including viewing screens, clicking, typing, navigating interfaces, and performing cross-tool operations.
Scientific and Technical Research: It outperforms other models in sustained capabilities within research loops, helping researchers explore ideas, gather evidence, test hypotheses, interpret results, and decide on next steps.
GPT-5.5 is now the default model in Codex. Users can refer to the API prompt guide and migration guide to fully leverage its advantages.