OpenAI explains when to use each GPT-6 model

OpenAI has published a practical guide for choosing and deploying models in the GPT-6 family, turning the launch into a clearer product map for builders.

The guide positions GPT-6 Astra as the option for the hardest reasoning work, GPT-6.1 Sol for complex coding, research and computer use, and GPT-6 Luna for focused repeated tasks at scale, such as extraction, classification and structured summaries. OpenAI also tells API users to tune reasoning effort by workload instead of treating model choice as the only lever.

The production advice is as important as the model list. OpenAI recommends using prompt caching and context compaction to manage cost, measuring success and latency per task, and setting up monitoring and data controls before deployment. For long-running agent work, it points builders toward mid-run steering, asynchronous tool calls and delegated multi-agent workflows.

For ToolCall readers, the useful signal is not a new benchmark. It is OpenAI making the GPT-6 family less mysterious: pick the model for the job, define what the agent can do independently, and design production workflows around cost, state and supervision.