本月早些时候，我们推出了 GPT‑6 Astra，这是世界上最智能、最对齐的模型。虽然最严苛和最重要的项目仍然需要 Astra 的全部深度，但工作会以不同的规模、节奏和预算发生。
这就是为什么我们通过 GPT‑6 Sol 和 GPT‑6 Luna 扩展 GPT‑6 宇宙。GPT‑6 Astra 引入了新一代智能——这些模型通过推进成本效率的前沿，帮助分配这种智能带来的益处。我们使用与 GPT‑6 Astra 类似的方法训练了 GPT‑6 Sol 和 Luna，将 Astra 在专业工作、事实性、编程、计算机使用和对齐方面最先进性能背后的进步，带到了更快、更实惠的模型中。
GPT‑6 模型在成本—智能曲线上领先，将每一层级的卓越能力与能够高效大规模交付这些能力的基础设施相结合。缓存和推理方面的改进让我们能够以更低的成本提供这些模型，我们通过将 Sol 和 Luna 的 API 价格相比其 GPT‑5.6 促销价格降低 50%，把这些节省直接传递给用户和客户。这些改进共同使先进 AI 对更多日常任务和大规模应用变得切实可行。
GPT‑6 API 定价
模型输入输出****价格降幅 GPT‑5.6 Sol → GPT‑6 Sol$4 → $2$20 → $10便宜 50% GPT‑5.6 Luna → GPT‑6 Luna$0.20 → $0.10$1.20 → $0.50便宜 50%
价格为每 100 万 tokens。
GPT‑6 Astra 仍然是我们在所有方面最好的模型。当你想要最佳结果和不妥协的体验时，请选择它。
整个模型家族的一次提升
GPT‑6 Sol 和 Luna 为你已经熟知并使用的模型带来了智能升级和成本效率，覆盖了对完成复杂工作最有用的能力。
专业工作
GPT‑6 Sol 可以承担困难的工作任务，同时通过更高的使用限制和更低的成本给你更多迭代空间，相比同等价位的竞品模型提供更多智能和更好的结果。
在 AutomationBench 上，这是一项跨应用测试业务工作流的基准，GPT‑6 Sol 在 xhigh effort 下以仅为 Claude Opus 5 在 max effort 下每任务成本 9% 的成本，超过了 Claude Opus 5。在高 effort 下，GPT‑6 Luna 相比其前代提升了 5.4 个百分点，每任务成本降低 58%。
在_AutomationBench 1.0.6_（在新窗口中打开）中，AI agents 在销售、营销、运营、支持、财务和人力资源领域的 47 个工具上接受端到端工作流测试。Claude Fable 5.1 的数据点低估了其实际成本，因为它省略了 Opus 5 fallback 的成本，而该 fallback 在约 40% 的任务中发生。_
GPT‑6 Sol 也以远低于 Claude Fable 5.1 的成本超过了它，甚至胜过低 effort 的 GPT‑6 Astra。
| 模型（及 effort） | 得分 | 每任务成本 |
|---|---|---|
| GPT‑6 Sol (xhigh) | 33.2% | $0.27 |
| GPT‑6 Astra (low) | 30.3% | 3.9 倍GPT‑6 Sol |
| Claude Opus 5 (max) | 26.9% | 11.1 倍GPT‑6 Sol |
| Claude Fable 5.1 含 Opus 5 Fallback (max) | 31.4% | >8.9 倍GPT‑6 Sol （fallback 成本未报告） |
在 Agents’ Last Exam 上，该基准评估 agents 在复杂专业工作流中的表现，GPT‑6 Sol 在 max effort 下得分为 56.4%，高于 Claude Opus 5 在该评估中的最高分，且每任务成本低 60%。
事实性
答案的有用性取决于事实是否正确，我们正在继续在事实可靠性方面取得进展。在我们的内部事实性评估中，该评估基于去标识化的真实世界对话，其中用户标记了我们模型的错误，GPT‑6 Sol 的错误数量约为其前代的一半，以低得多的成本接近 Astra 级别的可靠性。GPT‑6 Luna 也有显著提升；在更高 effort 水平下，它以约百分之一的成本匹配 GPT‑5.6 Sol。
这里我们在去标识化的 ChatGPT 对话上评估事实性，这些对话中用户曾标记先前模型的事实错误。这些诱发错误的对话并不代表典型使用情况，在典型使用中事实错误更为罕见。得分未按长度控制；然而，我们的冗长度扫描显示几乎不依赖答案长度。
编程
今年，编程 agents 开始处理比以往任何时候都更复杂、范围更大、持续时间更长的任务。在 OpenAI，我们的内部使用量呈指数增长。按 API 价格计算，中位数研究人员的每日 token 使用量已超过 $600，而第 90 百分位研究人员超过 $7,000（研究加速：OpenAI 内部视角）。随着编程 agents 承担更长、更苛刻的任务，持续使用的成本变得更加重要。GPT‑6 Sol 和 Luna 将强大的编程性能与更低的 API 价格相结合，为开发者提供更多迭代空间，也让团队更有信心对 Codex 提出更雄心勃勃的要求。
在 FrontierCode 上，该基准评估编程 agents 是否产生可合并到真实代码库中的更改，GPT‑6 Sol 相比 GPT‑5.6 Sol 有显著提升，并且能够以低得多的成本匹配 Claude Fable 5.1 xhigh。
在 DeepSWE v1.1 上，该基准测试在真实代码库中复杂软件工程任务上的表现，GPT‑6 Sol 在 max effort 下得分为 68.8%，与 Claude Fable 5 在该评估中的最高分——xhigh effort 下的 69.9%——相差 1.1 个百分点，而每任务成本约低 80%。
GPT‑6 Luna 在 max effort 下得分为 66.6%，与 Claude Opus 5 和 Fable 5 在 medium effort 下相当。在这些比较中，Luna 每任务成本比 Opus 5 低 93%，比 Fable 5 低 96%。
计算机使用
虽然 GPT‑6 Astra 仍然是世界上最好的计算机使用模型，但 GPT‑6 Sol 和 Luna 提供了比其前代更具成本效率的性能。在 OSWorld 2.0 offline 上，GPT‑6 Sol 在 xhigh effort 下取得了与 Claude Opus 5 在 medium effort 下相似的得分——60.5% 对 60.3%——而每任务成本约低 80%。GPT‑6 Luna (max) 能够以十分之一的成本超过 GPT‑5.6 Sol (medium)。
在_OSWorld 2.0_（在新窗口中打开）中，AI agents 尝试跨越日常和专业任务的长期计算机使用工作流。我们报告 v2026.08.08 版本离线集上的部分奖励。_
协作风格
我们还将 GPT‑6 Astra 改进后的沟通风格带到了 Sol 和 Luna，我们认为这会在技术和编程对话中尤其明显。可以期待更清晰、更少术语、更少奇怪的措辞、更少低价值细节，以及整体上略短的答案，同时不损失实质内容。
Prompt
Website’s looking clean! Could we use a Bento Box design style and add a slider between the pages in the top right? You may need to dive into the React..
GPT-5.6 Sol
GPT-6 Sol
尽管风格是主观的，但我们更偏好 GPT‑6 Sol 在此处的回复。它不会那么快下结论，花更少时间重复对提问者来说可能显而易见的细节（例如网站有四个页面），使用更少模糊语言（例如“便当感”、“围绕那个系统重塑”），更坦率地说明它检查了什么、没检查什么，并且不会不必要地分享其图像工具提示等实现细节。
改进智能体和长对话的缓存
在降低 token 价格的同时，我们正在帮助基于 GPT‑6 构建的开发者在其应用复用的上下文上节省更多。我们改进了 GPT‑6 的提示缓存，默认提供更高的缓存命中率，帮助智能体复用更多上下文、更快响应，并从缓存输入 token 读取 90% 的折扣中受益。
开发者也有更多方式来衡量和优化其缓存性能：
监控和诊断。 提示缓存仪表板（在新窗口中打开）显示有多少输入被缓存以及该情况如何随时间变化。诊断工具（在新窗口中打开）有助于解释错过的缓存机会以及需要修复的内容。
调整推理力度和工具可用性而不破坏缓存。 对更难的任务提高推理力度（在新窗口中打开），对更简单的后续任务降低它，并随着智能体需求变化启用或禁用工具（在新窗口中打开）。这两项控制现在都会保留较早的上下文以供缓存复用。
优化哪些前缀被缓存。 显式断点让开发者选择缓存提示前缀的结束位置。这使开发者对缓存复用有更多控制，并可以提升性能。
GitHub 报告称，在过去几个月里，这些改进已使 OpenAI 模型数十亿次请求中需要全新处理的提示 token 占比降低超过 50%，帮助 Copilot 更快响应。
持续改进对齐
GPT‑6 Sol 和 Luna 建立在 Astra 引入的对齐工作之上，Astra 是我们迄今为止对齐程度最高的模型。在我们的对齐评估中，Sol 和 Luna 都显示出相较其 GPT‑5.6 对应版本的改进，包括关于其编码工作的误导性声明比率更低。
以下评估有意测试具有挑战性的情境，并不衡量典型使用中的失败率。完整结果请参见系统卡（在新窗口中打开）。
在我们的内部编码欺骗评估中，AI 智能体被赋予刻意选择以诱发不诚实行为的任务。在典型使用中，欺骗要罕见得多。欺骗率衡量的是任何检测到欺骗的答案所占比例。力度设置为最大。
可用性
GPT‑6 Sol 和 GPT‑6 Luna 从今天起在 ChatGPT Work 和 Codex 中面向所有 Plus、Pro、Business、Enterprise 和 Edu 用户提供。Free 和 Go 用户可以在桌面应用中访问 GPT‑6 Luna。这些模型尚未在 Chat 中提供。在 OpenAI API 中，它们以 gpt-6-sol 和 gpt-6-luna 提供。
为了保持对所有人的服务稳定，我们计划在一天内逐步在 ChatGPT 中推出这些模型。如果你在 ChatGPT Work 或 Codex 中看不到新模型，请稍后再试。
GPT 的评估是在我们的研究环境中或通过我们的 API 进行的，由于系统提示、可用工具等差异，其输出可能与生产版 ChatGPT 略有不同。竞争对手模型的评估取自公开可用的报告。Claude Fable 5 的分数是在 Claude Fable 5.1 的分数不可用时报告的。
Earlier this month, we introduced GPT‑6 Astra, the most intelligent and aligned model in the world. While the most demanding and important projects still call for Astra’s full depth, work happens at different scales, rhythms, and budgets.
That’s why we’re expanding the GPT‑6 universe with GPT‑6 Sol and **GPT‑6 Luna.**GPT‑6 Astra introduced a new generation of intelligence—these models help distribute the benefits of that intelligence by advancing the frontier on cost efficiency. We trained GPT‑6 Sol and Luna with similar methods as GPT‑6 Astra, bringing the advances behind Astra’s state-of-the-art performance in professional work, factuality, coding, computer use, and alignment to faster, more affordable models.
The GPT‑6 models lead across the cost–intelligence curve, combining exceptional capabilities at every tier with infrastructure that delivers them efficiently at scale. Improvements in caching and inference let us serve these models at lower cost, and we’re passing those savings directly on to users and customers by reducing API prices for Sol and Luna by 50% compared with their GPT‑5.6 promotional pricing. Together, these improvements make advanced AI practical for more everyday tasks and applications at scale.
GPT‑6 API pricing
ModelInputOutput****Price reduction GPT‑5.6 Sol → GPT‑6 Sol$4 → $2$20 → $1050% cheaper GPT‑5.6 Luna → GPT‑6 Luna$0.20 → $0.10$1.20 → $0.5050% cheaper
Prices are per 1 million tokens.
GPT‑6 Astra continues to be our best model across the board. Choose it when you want the best results and an uncompromising experience.
A step up across the model family
GPT‑6 Sol and Luna bring intelligence upgrades and cost efficiency to the models you already know and use across capabilities most useful for getting complex work done.
Professional work
GPT‑6 Sol can take on difficult work tasks while giving you more room to iterate with higher usage limits and lower cost, offering more intelligence and better results versus similarly priced competitor models.
On AutomationBench, a test of business workflows across apps, GPT‑6 Sol at xhigh effort outperforms Claude Opus 5 at max effort at just 9% of Opus 5’s cost per task. At high effort, GPT‑6 Luna improves on its predecessor by 5.4 percentage points at 58% lower cost per task.
In_AutomationBench 1.0.6_(opens in a new window), AI agents are tested on end-to-end workflows using 47 tools across sales, marketing, operations, support, finance, and HR. The datapoint for Claude Fable 5.1 understates its actual cost, as it omits the cost of the Opus 5 fallbacks, which occurred on ~40% of tasks.
GPT‑6 Sol also exceeds Claude Fable 5.1 at far lower cost, and even bests low-effort GPT‑6 Astra.
| Model (and effort) | Score | Cost per task |
|---|---|---|
| GPT‑6 Sol (xhigh) | 33.2% | $0.27 |
| GPT‑6 Astra (low) | 30.3% | 3.9xGPT‑6 Sol |
| Claude Opus 5 (max) | 26.9% | 11.1xGPT‑6 Sol |
| Claude Fable 5.1 w/ Opus 5 Fallback (max) | 31.4% | >8.9xGPT‑6 Sol (fallback cost not reported) |
On Agents’ Last Exam, which evaluates agents on complex professional workflows, GPT‑6 Sol at max effort scores 56.4%, above Claude Opus 5’s highest score in the evaluation at 60% lower cost per task.
Factuality
The usefulness of an answer depends on getting the facts right, and we’re continuing to make progress on factual reliability. On our internal factuality evaluation, which is based on de-identified real-world conversations where users flagged mistakes by our models, GPT‑6 Sol makes about half as many mistakes as its predecessor, approaching Astra-level reliability at much lower cost. GPT‑6 Luna also improves substantially; at higher effort levels it matches GPT‑5.6 Sol at about a hundredth its cost.
Here we evaluate factuality on de-identified ChatGPT conversations where users had flagged a factual error from a prior model. These error-inducing conversations are not representative of typical usage, where factual errors are more rare. Scores are not controlled for length; however, our verbosity sweeps showed almost no dependence on answer length.
Coding
This year, coding agents have begun tackling tasks with more complexity, scope, and duration than ever before. At OpenAI, our internal usage has grown exponentially. Valued at API prices, daily token usage has exceeded $600 for the median researcher and $7,000 for researchers at the 90th percentile (Research acceleration: The view inside OpenAI). As coding agents take on longer and more demanding tasks, the cost of sustained use matters more. GPT‑6 Sol and Luna combine strong coding performance with lower API prices, giving developers more room to iterate and teams the confidence to be more ambitious about what they ask Codex to take on.
On FrontierCode, which evaluates whether coding agents produce changes ready to merge into real codebases, GPT‑6 Sol improves substantially over GPT‑5.6 Sol, and is able to match Claude Fable 5.1 xhigh at much lower cost.
On**DeepSWE v1.1,**which tests performance on complex software-engineering tasks in real codebases, GPT‑6 Sol at max effort scores 68.8%, within 1.1 percentage points of Claude Fable 5’s highest score in the evaluation—69.9% at xhigh effort—at approximately 80% lower cost per task.
GPT‑6 Luna at max effort scores 66.6%, comparable to Claude Opus 5 and Fable 5 at medium effort. In these comparisons, Luna costs 93% less per task than Opus 5 and 96% less than Fable 5.
Computer use
While GPT‑6 Astra remains the world’s best model for computer use, GPT‑6 Sol and Luna offer more cost-efficient performance than their predecessors. On OSWorld 2.0 offline, GPT‑6 Sol at xhigh effort achieves a similar score to Claude Opus 5 at medium effort—60.5% versus 60.3%—at approximately 80% lower cost per task. GPT‑6 Luna (max) is able to exceed GPT‑5.6 Sol (medium) at one tenth of its cost.
In_OSWorld 2.0_(opens in a new window), AI agents attempt long-horizon computer-use workflows spanning everyday and professional tasks. We report the partial reward on the offline set from the v2026.08.08 release.
Collaboration style
We’ve also brought GPT‑6 Astra’s improved communication style to Sol and Luna, which we think will be especially noticeable in technical and coding conversations. Expect to see more clarity, less jargon, fewer odd turns of phrase, fewer low-value details, and slightly shorter answers overall without losing substance.
Prompt
Website’s looking clean! Could we use a Bento Box design style and add a slider between the pages in the top right? You may need to dive into the React..
GPT-5.6 Sol
GPT-6 Sol
Although style is subjective, we prefer GPT‑6 Sol’s reply here. It doesn’t jump to conclusions as quickly, spends less time reiterating details that might be obvious to the asker (e.g., that the website has four pages), uses less vague language (e.g., “bento feel”, “reshaping… around that system”), is more forthcoming with what it did and didn’t check, and doesn’t unnecessarily share implementation details like its image tool prompt.
Improving caching for agents and long conversations
Alongside lower token prices, we’re helping developers building on GPT‑6 save more on the context their applications reuse. We’ve improved prompt caching for GPT‑6 to deliver higher cache hit rates by default, helping agents reuse more context, respond faster, and benefit from discounts of 90% on cached input-token reads.
Developers also have more ways to measure and optimize their caching performance:
Monitor and diagnose. The Prompt Caching Dashboard(opens in a new window) shows how much input is cached and how that changes over time. The diagnostics tool(opens in a new window) helps explain missed opportunities for caching and what to fix.
Adjust reasoning effort and tool availability without breaking cache. Increasereasoning effort(opens in a new window) for harder tasks or lower it for simpler follow-ups, andenable or disable tools(opens in a new window) as your agent’s needs change. Both controls now preserve earlier context for cache reuse.
Optimize which prefixes get cached. Explicit breakpoints let developers choose where cached prompt prefixes end. This gives developers more control over cache reuse and can improve performance.
GitHub reports that, over the past several months, these improvements have reduced the share of prompt tokens requiring fresh processing by more than 50% across billions of requests to OpenAI models, helping Copilot respond faster.
Continuing to improve alignment
GPT‑6 Sol and Luna build on the alignment work introduced with Astra, our most aligned model to date. In our alignment evaluations, both Sol and Luna show improvements over their GPT‑5.6 counterparts, including lower rates of misleading claims about their coding work.
The evaluations below deliberately test challenging situations and do not measure failure rates in typical use. See the system card(opens in a new window) for the full results.
In our internal coding deception evaluation, AI agents are given tasks deliberately selected to elicit dishonesty. In typical usage, deception is much rarer. Deception rate measures the fraction of answers with any detected deception. Effort was set to maximum.
Availability
GPT‑6 Sol and GPT‑6 Luna are available in ChatGPT Work and Codex starting today for all Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can access GPT‑6 Luna in the desktop app. These models are not yet available in Chat. In the OpenAI API, they are available as gpt-6-sol and gpt-6-luna.
To keep service stable for everyone, we plan to roll out these models in ChatGPT gradually throughout the day. If you don’t see the new models in ChatGPT Work or Codex, please try again later.
Evaluations of GPT were performed in our research environment or via our API, which may provide slightly different output from production ChatGPT due to differences in the system prompts, tools available, etc. Evaluations of competitor models were taken from publicly available reports. Scores for Claude Fable 5 were reported when scores for Claude Fable 5.1 were unavailable.
本文内容采集自官方网站，排版和翻译可能与原页面存在差异。
阅读官方全文