跳到正文
The Decoder· Jonathan Kemper·· 9 小時前AI 評分66

OpenAI 分享 GPT-6.2 Astra 與 Sol 的提示詞建議,改善答案過短及不必要追問

OpenAI shares new prompting tips for GPT-6.2 Astra and Sol to fix brief answers and unnecessary questions

AI 導讀

OpenAI 為 GPT-6.2 Astra 和 Sol 公佈可直接使用的提示詞範本,涵蓋寫作風格、主動執行任務及編碼方式。提示詞建議模型在獲授權時自行推進可逆步驟,令最終答案獨立完整;編碼時沿用現有邏輯,只有在提出要求、儲存庫規定或需防範特定回歸時才新增測試。

正文

OpenAI has released new prompt templates for GPT-6.2 Astra and Sol covering writing style, initiative, and coding, alongside API changes that can discard reasoning content.

OpenAI says developers moving to GPT-6.2 Astra or Sol should review their system prompts. The company published copy-ready prompts in the models' documentation to curb behaviors it observed in both releases. Astra is built for the most demanding tasks, while Sol handles complex work at lower cost.

Default conciseness changes how developers should prompt for writing

Astra writes more briefly and directly by default than earlier models, according to OpenAI. That means developers switching over should reconsider instructions like "be concise." For use as a colleague in everyday work, the company recommends a prompt that adopts user and source terminology, uses active voice, and puts key points first. When writing for external readers, the model should account for their lack of context. When editing, it should preserve the author's meaning and phrasing.

On long tasks, the model sometimes carries shorthand from its reasoning or other agents' messages into the final response. A dedicated prompt tells it to write that final answer so it stands on its own.

Treat your final answer as the user's first look at your work, and not as a continuation of intermediate commentary. The final answer must always be fully self-contained: users should never need to read earlier commentary updates for context, since they are collapsed when the final answer is shown to the user.

Write the final answer and all other human-facing material anew for the intended reader. Do not copy or closely paraphrase your reasoning or adopt its writing style. Do not carry over working shorthand, jargon, or abbreviations from your reasoning, tool calls, or instructions to subagents and automations.

GPT-6.2 should act on assumptions rather than wait for permission

When user intent is unclear, GPT-6.2 sometimes suggests a next step and waits for confirmation, OpenAI says. The company recommends a prompt telling the model to carry out those steps itself whenever it is authorized to do so. It should pause only for a major change in task scope, a real blocker, a decision the user must make, or an irreversible or destructive action.

When details are unclear, the model should make reversible assumptions, flag them, and keep working. Higher reasoning effort also helps it follow through on final checks during long tasks, OpenAI adds.

Avoid stopping to end the turn while the task is incomplete or if there is an authorized next step you can take, and instead take the next steps within task scope to progress the user's request. If your final response before ending the turn is a suggested next step, a plan, a request to the user for a routine decision, instructions for a task to execute, or similar handoff to the user, take these steps yourself to make progress towards what the user asked for. Make sure that you have authorization to do so, and flag any assumptions or decisions you make to the user.

Only stop to clarify with the user if there is significant change in task scope, you are concretely blocked, have a decision that the user needs to resolve, or before irreversible or destructive actions.

Leaner code output means fewer tests and less abstraction by default

OpenAI says GPT-6.2 writes more efficient code than its predecessors. Its template instructs the model to reuse existing logic and avoid unnecessary helper functions, while adding new tests only when requested, required by the repo, or needed to catch specific regressions. Fewer lines and shorter variable names are not a substitute for simplicity.

For frontends, developers should specify imagery and typography directly. By default, the template favors realistic images from image generation or search and uses SVGs only on request. Decorative numbering, filler badges, and helper text that only repeats what the interface already shows should be left out.

Above all, be thorough in producing high-quality, correct code that completely fulfills the user's task. Keep the code easy to read and verify that the user-requested outcome works as they intend. Follow these instructions when coding:

(a) Reuse existing logic in the user's repo when you can. You should reuse existing classes, methods, representations, and APIs in the codebase as long as they preserve correctness. Before consolidating or removing code, preserve the behavior its callers rely on unless the task requires changing it.
(b) Skip new tests unless requested, required by the repo, or needed to catch a specific regression existing tests would miss.
(c) Make sure that code is human readable. Do not use fewer lines, shorter variable names, minified formatting, or dense one-liners as substitutes for simplicity. Ensure that comments and variable names are readable and understandable to a human. Code-golfed expressions count as overly complex when they encode many branches, conditions, or hidden assumptions.
(d) Avoid unnecessary indirection and abstraction. Helper functions, utility functions, and class abstractions can add cognitive complexity even when they do not add branches.

Editing chat history can wipe earlier reasoning from the API

The API is changing too. Starting with GPT-6.2, the Responses API discards earlier reasoning content if developers edit conversation history, switch to certain other models, or change the end user or organization. That can hurt response quality and increase token usage as the model repeats work. For new integrations, OpenAI recommends the "medium" reasoning level, while existing setups using "none" or "minimal" should move to "low." Parameters like temperature and top_p aren't supported anymore.

In April, OpenAI advised against carrying over old prompt stacks for GPT-5.5 and urged developers to start fresh with short, outcome-focused prompts. September brought GPT-6 Astra prompts to curb frequent follow-up questions and ban AI clichés. OpenAI developer Eric Provencher also warned then that bloated skills and AGENTS.md files slow the model down. Back then, even small changes triggered thorough testing from Astra. GPT-6.2 is now meant to write new tests only on request.

OpenAI shipped GPT-6.2 Astra and Sol on November 17. Through the API, Astra costs $8 per million input tokens and $40 per million output tokens. Which model makes sense depends mainly on budget and quality needs, according to OpenAI employee Ted Sanders.

來源:The Decoder · the-decoder.com