OpenAI Launches Strongest New Model o1, Greatly Enhanced Reasoning Ability

OpenAI发布最强新模型o1，推理能力大幅增强

On September 12, OpenAI announced the launch of the new model o1-preview, seen as the preview version of its next-generation large model Orion.

9月12日，OpenAI宣布推出了新模型o1-preview，被视为其下一代大模型Orion（猎户座）的预览版本。

In the release document, OpenAI stated that the new model has reached a new level of artificial intelligence in terms of reasoning abilities, so the counter will be reset to 1, meaning there may not be a GPT-5 in the future.

OpenAI在发布文档中表示，新模型在推理能力上达到了人工智能的新水平，因此计数器将重置为1，这意味着未来可能不会有GPT-5。

Starting today, ChatGPT Plus and team users can directly access the model.

即日起，ChatGPT Plus和团队用户可以直接访问该模型。

Users can manually select to use either the o1-preview version or the smaller version o1-mini, with the former limited to 30 messages per week and the latter to 50.

用户可手动选择使用o1-preview预览版或小尺寸版本o1-mini，前者每周限制30条消息，后者限制50条。

The training data for the o1 model is current up to October last year.

在OpenAI的模型介绍页面上显示，o1模型的训练数据截止于去年十月。

Notably, early leaks about the Q* project appeared around November last year, matching this timeline.

值得注意的是，最早关于Q*项目的爆料大约出现在去年十一月，与该时间节点相符。

逐句精讲 | OpenAI发布史上最强新模型o1 06:09 Share

每日新闻 | 英语精讲

Shownotes Transcript

逐句精讲 | OpenAI发布史上最强新模型o1