To calibrate you all on which reason…

原文与翻译

原文（查看原帖）：

To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high.

If you were using high reasoning efforts with Sol and were happy, I suggest you move down to low or medium for Astra.

中文翻译：

【意译】To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high. If you were using high reasoning efforts with Sol and were happy, I suggest you move down to low or medium for Astra.

来源标注：@thsottiaux · 原帖链接

背景

公开资料中，GPT‑6 Astra / GPT‑6.1 Sol / GPT‑6.1 Terra / Ultrafast 等名称出现在 OpenAI DevDay 2026 与随后的产品更新（参见 DevDay 2026 Recap 与开发者 Changelog）。本帖是否对应某一具体正式发布，需以原帖时间线与官方文档为准；未点名的细节不作已核实事实。

意味着什么

Tibo 作为 Codex & ChatGPT 负责人，常在 X 上预告或确认模型侧变化。对订阅用户而言，关键点通常是：是否已在 ChatGPT / Codex / API 可选、用量如何计价、以及是否影响 dots / Work。

我的观点

模型名出现在高管时间线，往往早于或同步于帮助中心更新。中文用户建议同时核对 openai.com 与账户内模型选择器。