昨天，我们分享了GPT‑5.6如何帮助自身更高效地运行。今天，我们将这些成果传递给客户，通过降低GPT‑5.6 Luna(在新窗口中打开)和Terra(在新窗口中打开)的价格，以及提升API中GPT‑5.6 Sol的性能，让客户受益。这些更新共同帮助客户从投入AI的每一美元中获得更多价值，并在时间紧迫时更快行动。
从今天起，我们最快且最经济的模型GPT‑5.6 Luna将降价80%，而面向日常工作的均衡模型GPT‑5.6 Terra将降价20%。Luna和Terra的降价也反映在使用Codex和ChatGPT Work时，付费订阅的用量计算方式上。Luna为企业提供了一种更具成本效益的方式，以极高的质量处理大批量工作。它可以使用工具并完成多步骤工作流，使更广泛的AI应用能够大规模实际运行。
让先进智能更丰富、更实惠，是OpenAI确保AGI惠及全人类使命的核心。这些变化将这一承诺付诸实践。它们反映了我们在模型构建、服务和部署方式上多年来的改进。
我们还在API中推出了快速模式，取代了此前的优先处理服务。对于GPT‑5.6 Sol，快速模式现在比标准处理速度快达2.5倍，价格为两倍，智能水平不变。快速模式向后兼容：标记为优先的请求将自动使用快速模式。
1 / 6
将智能与结果匹配
高效使用AI始于结果。风险程度、错误成本、紧迫性和规模决定了智能、速度、可靠性和成本之间的正确平衡。这种平衡可能在工作流的每一步之间发生变化。
GPT‑5.6为企业优化这一等式提供了更大的空间。Luna提供的性能与一年前的尖端模型相当，而每项任务的成本约为其6%，速度则接近其九倍。根据Agents’ Last Exam的衡量，在专业工作上，Luna的表现优于Fable 5，而每项任务的估计成本低近99%。
在实践中，企业可以定义所需的结果和质量标准，然后使用评估来确定额外的智能在何处能显著改善结果，以及更快速、更低成本的处理在何处能提供相同的质量。例如，编码工作流可能使用Sol来解决不确定性并定义计划，然后使用Luna来实现明确指定的更改、编写和运行测试，并评估结果。另一个工作流可能需要不同的平衡。
GPT‑5.6系列扩展了这些选择的范围。企业可以在每个阶段应用最大限度的有用智能，同时为其创造的价值支付合适的价格。
实现这种灵活性始于让模型背后的每一层都更高效。
我们如何推进效率前沿
我们的效率优势来自于改进模型、运行模型的推理系统，以及将模型与工具和上下文连接起来的智能体框架。GPT‑5.6模型通过工作采取更直接的路径。更好的路由保持硬件高效运转，优化的生产软件更高效地生成令牌，更智能的上下文管理帮助智能体避免重复已完成的工作。这些改进共同使我们能够用相同的计算资源完成更多有用的工作，减少每个结果所需的时间、令牌和成本。
GPT‑5.6 Sol正越来越多地帮助我们找到并交付下一轮收益。在人类主导的流程中，Sol自主重写并优化了生产内核，设计并运行了数百次实验以改进令牌生成，并监控训练过程，在出现问题时进行干预。内核工作帮助将模型服务的端到端成本降低了20%，而其实验将令牌生成效率提高了超过15%。这项工作仍在继续，形成了一个更紧密的反馈循环：随着我们的模型改进并能够更自主地工作，我们提高效率的能力也在加速。了解更多关于GPT‑5.6背后的工程细节。
为规模化构建的计算策略
满足对丰富智能的需求需要更多的计算资源和更高效的计算资源。我们正在构建一个有弹性的基础设施组合，并将每个工作负载匹配到最适合运行它的系统。这种方法支持性价比曲线的两端。在低成本端，新的Luna和Terra价格使大批量工作在更大规模上变得经济实惠。在前沿端，快速模式为API客户在响应时间重要时提供更快的Sol访问。
企业可以将更多AI融入日常运营，而不会在最重要的工作上牺牲速度。大规模文档分析、客户互动分类和常规实施可以变得经济实惠地广泛运行，而复杂的Sol工作负载在溢价合理时可以更快运行。
收益可以累积。在人类主导的流程中，更强大的模型帮助我们的技术团队找到下一代的改进，缩短了通往更好性能和更低成本的道路。我们的策略仍然专注于推进能力和效率，使每一代智能都能以更低的成本完成更多工作。
可用性和定价
GPT‑5.6 Terra和Luna仍然在ChatGPT Work、Codex和OpenAI API中可用。在ChatGPT Work和Codex中，Free和Go用户可以使用Terra，而Plus、Pro、Business和Enterprise用户可以选择Terra和Luna。
从7月30日起，API定价为Terra每百万输入令牌2美元，每百万输出令牌12美元；Luna每百万输入令牌0.20美元，每百万输出令牌1.20美元。Sol定价保持不变。ChatGPT和Codex的订阅价格和配额预算保持不变，而Terra和Luna的使用现在消耗更少的积分。定价变更将于今天晚些时候开始在AWS中推出。
GPT‑5.6 Sol的快速模式取代了API中的优先处理，并与Codex中的/快速对齐。现有的标记为优先的API请求将继续工作。查看完整的API定价详情。
Yesterday, we shared how GPT‑5.6 helped make itself more efficient to run. Today, we’re passing those gains on to customers with lower prices for GPT‑5.6 Luna(opens in a new window) and Terra(opens in a new window) and faster performance with GPT‑5.6 Sol in the API. Together, these updates help customers get more from every dollar they invest in AI and move faster when time matters.
Starting today, GPT‑5.6 Luna, our fastest and most affordable model, will cost 80% less, while GPT‑5.6 Terra, our balanced model for everyday work, will cost 20% less. These lower prices for Luna and Terra are also reflected in how usage is counted against paid subscriptions when using Codex and ChatGPT Work. Luna gives businesses a far more cost-effective way to handle high-volume work at very high levels of quality. It can use tools and complete multi-step workflows, making a broader range of AI applications practical to run at scale.
Making advanced intelligence more abundant and affordable is central to OpenAI’s mission to ensure AGI benefits all of humanity. These changes put that commitment into practice. They reflect years of improvements in how our models are built, served, and put to work.
We’re also introducing Fast mode in the API, which replaces our Priority Processing offering. For GPT‑5.6 Sol, Fast mode now delivers up to 2.5× faster speeds than Standard processing at twice the price, with no change in intelligence. Fast mode is backward compatible: requests tagged priority will automatically use Fast mode.
1 of 6
Matching intelligence to the outcome
Using AI efficiently begins with the outcome. The stakes, cost of error, urgency, and scale determine the right balance of intelligence, speed, reliability, and cost. That balance can change from one step of a workflow to the next.
GPT‑5.6 gives businesses much more room to optimize that equation. Luna delivers performance comparable to models that were frontier-class a year ago at roughly 6 cents on the dollar per task, and at nearly nine times the speed. On professional work, as measured by Agents’ Last Exam, Luna outperforms Fable 5 at an estimated cost per task nearly 99% lower.
In practice, businesses can define the outcome and quality standard they need, then use evaluations to determine where additional intelligence materially improves the result and where faster, lower-cost processing can deliver the same quality. A coding workflow, for example, might use Sol to resolve uncertainty and define the plan, then use Luna to implement well-specified changes, write and run tests, and evaluate the results. Another workflow may call for a different balance.
The GPT‑5.6 family expands the range of those choices. Businesses can apply the maximum useful intelligence at every stage while paying the right price for the value it creates.
Delivering that flexibility starts with making every layer behind the models more efficient.
How we advance the efficiency frontier
Our efficiency edge comes from improving the models, the inference systems that run them, and the agentic harness that connects them to tools and context. GPT‑5.6 models take a more direct path through work. Better routing keeps hardware productive, optimized production software generates tokens more efficiently, and smarter context management helps agents avoid repeating completed work. Together, these improvements let us complete more useful work with the same compute, reducing the time, tokens, and cost required for each result.
GPT‑5.6 Sol is increasingly helping us find and deliver the next round of gains. Within a human-led process, Sol autonomously rewrote and optimized production kernels, designed and ran hundreds of experiments to improve token generation, and monitored training, intervening when problems arose. The kernel work helped reduce the end-to-end cost of serving the model by 20%, while its experiments increased token-generation efficiency by more than 15%. This work continues, creating a tighter feedback loop: as our models improve and are able to work more autonomously, our ability to improve efficiencies accelerates. Read more about the engineering behind GPT‑5.6.
A compute strategy built for scale
Meeting demand for abundant intelligence requires both more compute and more productive compute. We are building a resilient infrastructure portfolio and matching each workload to the systems best suited to run it. That approach supports both ends of the price-performance curve. At the lower-cost end, the new Luna and Terra prices make high-volume work economical at much greater scale. At the frontier end, Fast mode gives API customers faster access to Sol when response time is important.
Enterprises can move more AI into everyday operations without sacrificing speed on their most consequential work. Large-scale document analysis, customer-interaction classification, and routine implementation can become economical to run broadly, while complex Sol workloads can move faster when the premium is justified.
The gains can compound. Within a human-led process, more capable models help our technical team find the next generation of improvements, shortening the path to better performance and lower costs. Our strategy remains focused on advancing both capability and efficiency so each generation of intelligence can accomplish more work at a lower cost.
Availability and pricing
GPT‑5.6 Terra and Luna remain available in ChatGPT Work, Codex, and the OpenAI API. In ChatGPT Work and Codex, Free and Go users can access Terra, while Plus, Pro, Business, and Enterprise users can choose Terra and Luna.
Starting July 30, API pricing is $2 per million input tokens and $12 per million output tokens for Terra, and $0.20 per million input tokens and $1.20 per million output tokens for Luna. Sol pricing remains unchanged. ChatGPT and Codex subscription prices and quota budgets remain unchanged, while Terra and Luna usage now consumes fewer credits. Pricing changes will begin rolling out in AWS later today.
Fast mode for GPT‑5.6 Sol replaces Priority Processing in the API and aligns with /fast in Codex. Existing API requests tagged priority will continue to work. View complete API pricing details.
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