The builder’s guide to GPT‑5.6

GPT-5.6 is a new AI model family that achieves better performance at lower costs by reducing the number of tokens (small units of text that AI models process) needed for complex tasks. The guide explains that smaller models in the GPT-5.6 family (Luna and Terra) can now match the performance of more expensive, more powerful models, and introduces three new API features that help AI agents (software that performs tasks automatically) work more efficiently: reasoning persistence (keeping track of previous work across steps), native compaction (summarizing long conversations to save space), multi-agent orchestration (running multiple AI agents in parallel), and programmatic tool calling (having the AI delegate routine work to code instead of processing it internally).