Event ·Multi-source

微软发布首个网络安全模型MAI-Cyber-1-Flash，基准得分96%

微软发布了其首个网络安全模型MAI-Cyber-1-Flash，在CyberGym漏洞基准测试中取得96%的分数。该分数比Anthropic的Mythos高12个百分点，且推理成本仅为后者的一半。微软AI CEO Mustafa Suleyman表示，token成本已成为防御方的真正约束。

Current read

微软的新安全模型MAI-Cyber-1-Flash，比Anthropic Mythos高出12个点，成本还减半，搞网络安全的可以盯着。

11 sources78° AI heatLast updated 2026/7/27 17:22:22

Evidence

11 sources

Primary sourceThe Rundown AI

2026/7/27 17:22:22View source

Related source 1IT之家

2026/7/28 00:59:36View source

Related source 2orange.ai

2026/7/26 22:05:17View source

Related source 3Lenny Rachitsky

2026/7/27 21:32:57View source

Related source 4Decoder

2026/7/28 19:12:27View source

Related source 5SuperTechFans

2026/7/27 00:12:16View source

Related source 6AI Will

2026/7/27 02:52:52View source

Related source 7Thomas Wolf

2026/7/27 03:07:13View source

Related source 8Fireworks AI

2026/7/27 15:30:35View source

Related source 9Y Combinator

2026/7/27 17:02:38View source

Related source 10arXiv: Anthropic

2026/7/27 19:43:41View source

Conflict 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.

Open the story