# OpenAI publishes prompt templates for GPT-6.2 Astra and Sol and API changes

published: 2026-11-21T21:05:28.715Z
topic: infrastructure
source: https://the-decoder.com/openai-shares-new-prompting-tips-for-gpt-6-2-astra-and-sol-to-fix-brief-answers-and-unnecessary-questions/
url: https://aiagentnews.io/stories/openai-publishes-prompt-templates-for-gpt-6-2-astra-and-sol-and-api-changes

OpenAI has released copy-ready prompt templates for GPT-6.2 Astra and Sol covering writing style, initiative and coding. It also flagged API changes in which editing conversation history can discard earlier reasoning content.

OpenAI has published prompt templates for GPT-6.2 Astra and Sol in the models' documentation, covering writing style, initiative and coding. The company says both models answer more briefly by default than earlier versions, so developers should revisit instructions such as "be concise." The writing template asks for key points first, active voice and the terminology of the user.

The initiative template tells the models to carry out authorized next steps rather than end a turn with a suggestion or a routine request for approval. They should pause only for major scope changes, real blockers, decisions the user must make, or irreversible actions, and should flag any assumptions they make. The coding template asks for existing logic to be reused, unnecessary helper functions avoided, and new tests added only when requested or required.

On the API side, editing conversation history causes the Responses API to discard earlier reasoning content, which can raise token usage and reduce response quality. Parameters such as temperature and top_p are no longer supported. OpenAI recommends medium reasoning effort for new integrations and low for existing setups that use none or minimal.

## Why it matters if you build agents

Agent loops that rewrite message history may lose reasoning state and pay more tokens, so history edits should be kept to a minimum. Prompts written for earlier models, especially those demanding brevity or constant confirmation, may need to be rewritten. Code that sets temperature or top_p should drop those parameters before moving to GPT-6.2.
