模型多源确认
SemiAnalysis 发布 GDPx 基准，最强模型仅得 12%
精选理由
SemiAnalysis 搞了个新基准 GDPx，最强模型才拿 12 分，想看各家智能体真实水平差距的可以瞅瞅这份榜单。
SemiAnalysis 发布 GDPx 基准，用于测量模型在高价值领域的智能体能力。该基准难度极高，目前表现最好的模型得分只有 12%。作者还提到从 Argon 到 Carbon 的性能提升，以及 Gemini 模型在同等性能下的成本优势。
原文 · Varun Mohan
Always great to see new evals where the highest score is less than 15%. Also, cool to see the jump from Argon to Carbon and how cost effective the Gemini models are for their level of performance. Excited to see how this leaderboard looks like in a couple months given the rate of progress. SemiAnalysis @SemiAnalysis_ Today we’re releasing GDPx, a benchmark for measuring agentic capabilities across the most valuable domains. GDPx is a true frontier-level benchmark: even the best model scores just 12% (1/7)🧵 🔗 View Quoted Tweet 💬 35 🔄 9 ❤️ 310 👀 20270 📊 61 ⚡
