OpenAI publishes prompt templates for GPT-6.2 Astra and Sol, and changes Responses API reasoning handling
The templates cover writing, initiative and coding. Astra costs $8 per million input tokens and $40 per million output tokens through the API, and both models shipped November 17.
What happened
OpenAI gave developers ready-made instructions, called prompts, for its new GPT-6.2 Astra and Sol models. The prompts adjust how the models write, how often they stop to ask questions, and how they write code. Astra is aimed at the hardest tasks, and Sol handles complex work at lower cost.
Why it matters to you
Anyone building tools on these models may see shorter answers and fewer check-in questions unless they update their instructions. Software teams could also get less test code by default.
The takeaway
If you built products on older OpenAI models, your old instructions may need a rewrite for GPT-6.2.
What it means for you
Audit system prompts, remove temperature and top_p settings, and avoid editing conversation history mid-session so reasoning content is not discarded.
Teams migrating to GPT-6.2 should budget time to retest prompts, and pick Astra or Sol based on cost and quality needs, as OpenAI's Ted Sanders says.
OpenAI lists Astra API pricing at $8 per million input tokens and $40 per million output tokens; the models shipped November 17.