Primary sourceIT之家
View sourceEvent ·Multi-source
OpenAI GPT-5.6 IQ 首破 130 天才线,超越 99% 人类
在 Tracking AI 最新离线 IQ 测试中,GPT-5.6 全系列模型(SOL、TERRA、视觉版等)均获得 136 分,首次将大模型智商推过 130 的天才门槛,全球仅约 1% 人类达到此分数。此前 o3 等旗舰模型均卡在 130 分以下,Claude-5 Fable 以 130 分紧随其后。开发者测试显示 GPT-5.6 Sol 能通过单条 prompt 生成物理模拟、RAG 客服系统等应用,成本仅为 Fable 5 的零头。但 IQ 测试仅衡量抽象模式识别和逻辑推理,无法评估事实可靠性、工具调用等真实场景能力。
Current read
GPT-5.6 智商飙到 136 分,比 Claude-5 还高出一截,而且干活也不含糊,一句话就能造出应用,值得看看它到底多强。
11 sources41° AI heatLast updated 2026/7/16 02:51:14
Evidence
11 sourcesRelated source 1Greg Brockman
View sourceRelated source 2Sam Altman
View sourceRelated source 3OpenAI
View sourceRelated source 4shao__meng
View sourceRelated source 5歸藏(guizang.ai)
View sourceRelated source 6AI Will
View sourceRelated source 7@OpenAIDevs
View sourceRelated source 8Decoder
View sourceRelated source 9elvis
View sourceRelated source 10arXiv: OpenAI
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.