本月早些时候，OpenAI 推出了 GPT-6 Astra，并将其誉为迄今为止最强大、能力最强的模型，以及在包括计算机工作和编程在内的多种活动中的“全球最佳模型”。如今，该实验室正在通过更新较小规模的 Sol 和 Luna 模型版本来扩展 GPT-6 系列。

该公司在声明中写道：“GPT-6 Astra 引入了一代新的智能；这些模型通过使这种智能更加高效和易于访问，将其益处进一步延伸。”

Sol 和 Luna 系列的首批模型于今年早些时候推出，代表了 OpenAI 模型层级中的不同档次。Sol 专为编码等复杂任务而设计，而 OpenAI 表示 Luna 更适合文书工作——即“具有明确目标的高容量任务，例如总结文档、提取信息或回答快速问题。”

OpenAI 还强调了在效率和成本效益方面的重大改进——最显著的是 API 访问价格的大幅下调。6 系列模型的价格将是 Sol 和 Luna 5.6 系列的一半，OpenAI 将此价格下降归因于缓存和推理技术的改进。

OpenAI 还声称，其最新模型在生成事实准确的答案方面表现更佳，并指出它们在编程方面的错误率更低。

公告称：“在我们基于去标识化的真实世界对话进行的内部事实性评估中（用户在这些对话中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，以低得多的成本达到了 Astra 级别的可靠性。”

与 OpenAI 发布新模型的一贯做法相同，该公司声称其最新产品优于其主要竞争对手 Anthropic 的产品。该公司多次声称，GPT-6 Sol 和 Luna 处理任务的能力显著优于 Anthropic 的顶级模型——如 Fable 和 Opus。

值得注意的是，Anthropic 在 OpenAI 发布前仅 90 分钟就发布了 Opus 5.5 的新版本，这反映了两家公司之间激烈的竞争态势。

新版 Sol 和 Luna 现已面向大多数付费账户在 ChatGPT Work 和 Codex 中可用，并可通过 ChatGPT API 访问；而 Luna 将在桌面应用程序以及免费用户和 Go 版用户中提供。OpenAI 预计将在今天内逐步将模型推向 ChatGPT（应用程序和网站）。

Earlier this month, OpenAI launched GPT-6 Astra , which it heralded as its most powerful and capable model yet, and the “world’s best model” for a variety of activities, including computer work and coding. Now, the lab is expanding the GPT-6 generation with updated versions of the smaller Sol and Luna models.

“GPT-6 Astra introduced a new generation of intelligence; these models extend its benefits by making that intelligence more efficient and accessible,” the company writes.

The first models in the Sol and Luna series were introduced earlier this year , representing different tiers in OpenAI’s hierarchy of models. Sol is designed for complex tasks like coding, while OpenAI says Luna is more appropriate for clerical work — “high-volume tasks with a clear goal, like summarizing documents, extracting information, or answering quick questions.”

OpenAI is also emphasizing big improvements in efficiency and affordability — most notably, via a substantial price drop for API access. The 6 series models will be available at half the cost of the 5.6 series of Sol and Luna, a price drop that OpenAI attributes to improvements in caching and inference.

OpenAI also claims that its newest models are better at producing answers that are factually accurate, and notes that they have a lower error rate when it comes to coding.

“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, reaching Astra-level reliability at much lower cost,” the announcement reads.

As is typical with OpenAI’s new model releases, the company is claiming that its newest offerings outshine those of its primary competitor, Anthropic. The company repeatedly claims that GPT-6 Sol and Luna handle tasks substantially better than Anthropic’s top models — like Fable and Opus.

Notably, Anthropic released a new version of Opus 5.5 just 90 minutes before OpenAI’s release, reflecting the intense competition between the two companies.

The new versions of Sol and Luna are now available in ChatGPT Work and Codex for most paid accounts, and in the ChatGPT API, while Luna will be available in the desktop app and for Free and Go users. OpenAI expects to roll out the models to ChatGPT (the app and website) gradually throughout the day.