OpenAI的下一代主要模型Astra的内部版本成功解决了十个数学、量子复杂性和理论计算机科学领域的重大开放问题[1]。根据EpochAI Research的评估标准，这些成果的重要程度经过了系统分类和评价。
在对这些突破进行的评估中，通过GPT-5.6 Sol Pro和Fable 5 Max两个模型的检验结果显示，问题#3被双方都认定为"突破性进展"[1]。整体而言，至少七项成果被评定为"重大进展"[1]，而Fable 5 Max还将问题#1、#4和#9认定为"接近突破性"的成果[1]。
OpenAI's next-generation flagship model, Astra, has resolved ten significant open problems spanning mathematics, quantum complexity, and theoretical computer science in its internal versions.[1] The breakthrough solutions were evaluated using assessment criteria developed by EpochAIResearch's OpenMath standards and analyzed through two advanced models: GPT-5.6 Sol Pro and Fable 5 Max.[1]
According to the evaluation framework, problem number three received the highest classification of "breakthrough" from both models, marking it as a particularly significant achievement.[1] At minimum, seven of the ten solutions were rated as "major advances," demonstrating substantial progress across multiple research domains.[1] Additionally, Fable 5 identified at least three problems—numbered one, four, and nine—as "borderline breakthrough" contributions, suggesting particularly promising developments at the frontier of these research areas.[1]