OpenAI
GPT-6.2 Astra
最大上下文
1.1M
最大输出
128K
发布日期
暂无
gpt-6.2-solGPT-6.2 Sol 是 OpenAI GPT-6.1 Sol 的升级版,在 GPT-6.2 系列中定位低于旗舰 GPT-6.2 Astra。它是一款高效、通用的日常模型,适用于多步骤智能体工作流,以及技术与非技术任务中清晰自然的写作。
上下文窗口
1.1M Token
最大输出
128K Token
发布日期
2026年11月17日
模态
| Token 阶梯 | 输入价格 | 输出价格 | 缓存读取 | 缓存创建 5m |
|---|---|---|---|---|
| <=272K | $2/M | $10/M | $0.1/M | $2.5/M |
| >272K | $4/M | $15/M | $0.2/M | $5/M |
推理
工具调用
Temperature 参数
附件
知识库
Endpoint 协议
GPT-6.2 Sol 为日常软件开发、写作和经常运行的工作流提供推理能力。
能处理各步骤相互依赖的任务,适合需要经常运行、关注 Token 成本的工作流。
能结合现有代码和测试进行分析,适合日常功能实现和代码审查。
能起草和修改专业文稿,表达清楚,并根据读者调整语气。
能把文字说明与截图、示意图或其他参考图片里的细节结合起来分析。
用 GPT-6.2 Sol 处理日常代码修改、可重复的流程和需要审阅的文稿。
修复范围明确的缺陷或实现小功能,更新相关测试,并为审查者总结改动。
在已连接的系统中查询工单,补齐所需信息,再生成结构化交接记录供下一团队处理。
根据拉取请求摘要和产品记录,整理成适合目标读者的发布说明或分步操作指南。
把截图与预期行为对照,说明可见差异,并整理出供测试人员复现的步骤。
import OpenAI from "openai"
const client = new OpenAI({
apiKey: process.env.TOKENHUB_API_KEY,
baseURL: "https://us-api.tokenhub.com/v1",
})
const result = await client.responses.create({})
console.log(result.output_text)了解 GPT-6.2 Sol 的上下文限制、输入类型、标准价格和 API 接入方式。
它采用 OpenAI 面向日常工作的 Sol 推理模型,适用于软件开发、写作和需要多个步骤的工作流。
OpenAI 公布的上下文窗口为 105 万 Token,单次最多输入 92.2 万 Token、输出 12.8 万 Token。
OpenAI 模型可接收文本和图片,输出为文本。音频和视频不在原生支持范围内。
使用 TokenHub API 密钥,把 model 字段设为本页显示的模型 ID。所选端点的基础 URL 和请求格式请查看 API 接入部分。
输入不超过 27.2 万 Token 时,OpenAI 标准价格为每百万 Token 输入 2 美元、缓存读取 0.10 美元、输出 10 美元。超过该门槛时,整次请求按长上下文档位计费。TokenHub 价格以本页价格表为准。
日常编码、文稿起草和重复性工作流可以先用 Sol,兼顾成本与响应速度。涉及复杂的跨模块修改或高难度视觉任务,且预算允许时,可以评估 Astra。
通过一个 API Key 接入 GPT-6.2 Sol 与更多 AI 模型。
GPT-6.2 Sol 视频与社区讨论
收录 GPT-6.2 Sol 的发布帖、实测演示和开发者讨论。
X (Twitter)
OpenAI ·
GPT-6.2 Sol—an efficient and versatile model for everyday work. It powers through multistep agentic workflows, writes clearly and naturally, and is intuitive to collaborate with. Many OpenAI employees use GPT-6.2 Sol as their go-to model across technical and non-technical work. https://t.co/fOhGkAbgjb
https://pbs.twimg.com/media/HXWC7lJboAAhE-y.png查看原文OpenAI ·
Some of you have been asking very politely. Some of you have just been asking. GPT-6.2 Astra and Sol are now available. Super efficient and built to perform where it counts—the real world. https://openai.com/index/introducing-gpt-6-2-astra-sol https://t.co/jLCyVIws2Y
https://pbs.twimg.com/media/HXbQg3xacAAib6t.jpg查看原文OpenAI Developers ·
Upgrades, people. Upgrades. The wait is over, you can now use GPT-6.2 Astra and GPT-6.2 Sol in Codex and the API. https://t.co/Uz6JF5XAF1
https://pbs.twimg.com/media/HXWMJzlWQAA4bE2.jpg查看原文Reddit
then, how is your experience with 6.2 Astra and 6.2 Sol? megathread
Lucas_Zxc2833 ·
so, wanting to do this post to know how your experiences with Chatgpt are going, if everything is going well, if they are going bad, I’ve been hearing and seeing both good and bad things, but I wanted to know everything in this post share everything you have to share here, also tips and tricks on how to use it better I haven’t tested yet, but soon I plan
查看原文GPT-6.2 Sol Just Dropped and Users Are Already Fighting Over It
Optimum1997 ·
查看原文Sol 6.1 vs. 6.2 in Codex: Early results with updated AGENTS.md
Finkk1 ·
展开全文
TL;DR: Sol 6.1 worked perfectly for me for strategy and message drafts. So far, 6.2 hasn’t felt as good for that work, but I love how fast it is, especially in Ultrafast. I’ll test it for a few days with updated status files and see whether it can match 6.1’s quality. Sol 6.2 on Max also recommended switching to 6.1 without me asking. Some of the difference may come from stale status files and overlapping instructions. I cleaned up the instructions and added a short rule to my global AGENTS.md: evaluate the move I asked about, separate confirmed facts from assumptions, and include the proposed next step in the draft itself. Model Effort AGENTS.md rule Score Sol 6.1 Medium Old 9/12 Sol 6.2 Medium Old 9/12 Sol 6.2 High Old 8/12 Sol 6.1 Medium Revised 12/12 Sol 6.2 Medium Revised 12/12 Both models tied with the revised rule. That was one run per setup across four scenarios. The next test is whether 6.2 holds up or outperforms 6.1 in actual work over several days.
YouTube
GPT-6.2 Astra and GPT-6.2 Sol—Models built for the real world
OpenAI
查看原文GPT-6.2 Astra vs Sol | First Impressions
Arena
查看原文【GPT-6.2 Sol】ChatGPT新モデルを4つの仕事で比較|スライド作成・旅行の計画・売上グラフ・3D
AI market
查看原文