Mashable 06月11日 23:51
OpenAI launches new, smarter model. Meet o3-pro.
index_new5.html
../../../zaker_core/zaker_tpl_static/wap/tpl_guoji1.html

 

OpenAI发布了最新的推理模型o3-pro,声称其在科学、教育、编程、数据分析和写作方面表现更出色。o3-pro已向ChatGPT Pro和Team用户开放API,并将在下周向企业版和教育版用户推出。该模型通过将任务分解为多个步骤来提高答案的准确性和可靠性。尽管如此,苹果的研究人员发现,在解决特定问题时,如汉诺塔问题,推理模型仍存在局限性。o3-pro在ChatGPT中可以使用网络搜索、Python、文档视觉分析等工具,但不支持图像生成和画布功能。由于使用了工具,o3-pro的响应时间可能会更长。

🧠o3-pro是OpenAI最新的推理模型,在科学、教育、编程、数据分析和写作方面表现更优。

💡o3-pro通过将任务分解为步骤来提高答案的准确性和可靠性,与传统的LLM有所不同。

🛠️o3-pro向ChatGPT Pro和Team用户开放API,并将在下周向企业版和教育版用户推出。它可以使用ChatGPT中的多种工具,如网络搜索和Python。

⚠️尽管o3-pro有所改进,但苹果的研究人员发现,在解决如汉诺塔问题等特定问题时,推理模型仍存在局限性。

OpenAI has a new reasoning model called o3-pro that the company says is its most intelligent yet.

On Tuesday the ChatGPT maker announced o3-pro on X, sharing some details on its improvement over o3. OpenAI highlighted better performance in "science, education, programming, data analysis, and writing" and also said reviewers rated it higher on "clarity, comprehensiveness, instruction-following, and accuracy."

o3-pro is available in the API and to ChatGPT Pro and Team users, with Enterprise and Edu availability rolling out next week.

OpenAI's o3-pro is the newest addition to its family of reasoning models, which break down tasks into steps for ostensibly more accurate and reliable responses, as opposed to conventional LLMs. In this regard, reasoning models like o3-pro are considered better for complex tasks and OpenAI says o3-pro "excels at math, science, and coding." OpenAI posted benchmark evaluations indicating o3-pro surpasses o1-pro and o3 in these areas.

Researchers from Apple recently found some notable limitations in reasoning models, including OpenAI's o3-mini. When prompted to solve the classic Tower of Hanoi problem, which involves moving discs from one peg to another, the models struggled as complexity increased and even gave up despite having more computing power at their disposal.

The research made waves amongst the AI community as evidence that reasoning models aren't as smart as they're hyped to be. However, experts also pointed out that the research tested a very specific problem, and it didn't compare the results of humans and reasoning models on the same problems. So, it's not a definitive conclusion that AI tools are completely overhyped, but the findings do suggest that they're not necessarily the best models for every kind of task.

o3-pro has access to tools in ChatGPT, including web search, Python, visual analysis of documents, and personalized responses with memory support. But it doesn't support image generation or canvas, OpenAI's interface for working on projects, according to the release notes. Responses will also take longer because o3-pro has access to tools.


Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.

Fish AI Reader

Fish AI Reader

AI辅助创作,多种专业模板,深度分析,高质量内容生成。从观点提取到深度思考,FishAI为您提供全方位的创作支持。新版本引入自定义参数,让您的创作更加个性化和精准。

FishAI

FishAI

鱼阅,AI 时代的下一个智能信息助手,助你摆脱信息焦虑

联系邮箱 441953276@qq.com

相关标签

OpenAI o3-pro 推理模型 人工智能
相关文章