OpenAI Publishes GPT-6 Family Guide Covering Model Choice, Long-Running Work, and Production

OpenAI has published “A model guide for the GPT‑6 family,” a practical guide to balancing capability, cost and latency across the GPT‑6 lineup. OpenAI positions GPT‑6 Astra for the hardest reasoning work, GPT‑6.1 Sol for complex coding, research and computer use, and GPT‑6 Luna for focused, high-volume tasks with clear goals. The guide also explains how to tune reasoning effort and when to consider Fast or Ultrafast speed modes.

For production workloads, OpenAI recommends trimming unnecessary context, reusing stable material with prompt caching, and using compaction to manage long conversations. Teams are advised to test representative tasks before deployment and measure task success, latency and cost per successful task. The guide also calls for reviewing prompts, skills and AGENTS.md files so the model has explicit boundaries around what it may do independently, when it should ask for input and what counts as complete.

For long-running agent work, the guide covers mid-turn steering, asynchronous tools and parallel or delegated work, along with computer use for interacting with websites and desktop applications. The overall guidance is to match model and reasoning level to the workload, then support the system with monitoring, data controls, caching and clear operational boundaries.