A model guide for the GPT-6 family
This guide explains how to choose among GPT-6 models, prompt them effectively, manage long-running work, and deploy them in production.Summary
OpenAI's GPT-6 family offers three tiers: GPT-6 Astra for the hardest reasoning, GPT-6.1 Sol for complex coding and research, and GPT-6 Luna for focused tasks at scale. The guide advises matching model and reasoning level to the workload, using Fast mode or Ultrafast when speed matters, and estimating cost with caching. For prompts and skills, it recommends clear assignments, concise skill descriptions, updated AGENTS.md instructions, explicit decision boundaries, and prescriptive definitions of "done". For long-running tasks, the API supports mid-turn steering, asynchronous tool calling, and multi-agent delegation, while Codex can ask clarifying questions and be redirected as requirements change. Computer use lets models interact with websites and desktop apps. Before production, teams should test representative tasks, measure success and latency, use prompt caching and compaction, and review monitoring and data controls.
(Source：OpenAI)