3 Matching Annotations
  1. Apr 2026
  2. www.minimax.io www.minimax.io
    https://www.minimax.io/models/text/m27
    1
    1. fxp007 16 Apr 2026
      On GDPval-AA, M2.7 achieves an ELO score of 1495, the highest among open-source models.

      令人惊讶的是：MiniMax M2.7在GDPval-AA基准测试中获得了1495的ELO分数，成为所有开源模型中的最高分。这一分数不仅展示了模型在专业办公领域的卓越能力，还暗示了开源AI模型已经达到了接近或超越某些专有模型的专业水平，打破了开源模型性能不如闭源模型的刻板印象。

      surprising benchmark-results
    Visit annotations in context

    Tags

    Annotators

    URL

    minimax.io/models/text/m27
  3. www.theaivalley.com www.theaivalley.com
    https://www.theaivalley.com/p/the-claude-mythos-era
    1
    1. fxp007 16 Apr 2026
      The model reportedly scored 93.9% on SWE-bench Verified and 77.8% on SWE-bench Pro, but its strongest signal came from real-world results, including uncovering a 27-year-old flaw in OpenBSD, a 16-year-old vulnerability in FFmpeg, and autonomously chaining Linux kernel exploits without human input.

      这些惊人的安全漏洞发现能力表明AI已经超越了传统安全工具，能够自主发现几十年未被发现的漏洞。特别是能够自主链接Linux内核漏洞的能力，展示了AI在网络安全领域的革命性潜力，这可能彻底改变安全研究和漏洞修复的方式。

      ai-security benchmark-data real-world-results
    Visit annotations in context

    Tags

    Annotators

    URL

    theaivalley.com/p/the-claude-mythos-era
  4. arxiv.org arxiv.org
    https://arxiv.org/abs/2604.02947
    1
    Visit annotations in context

    Tags

    Annotators

    URL

    arxiv.org/abs/2604.02947