Claude Opus 5（自适应推理、最大努力模式）在Artificial Analysis Intelligence Leaderboard上获得第一名 [1]。该模型在满分170个模型评估中得分61分 [1]。
Artificial Analysis Intelligence Leaderboard对586个模型进行了多维度对比评估 [1]，涵盖智能程度、定价、输出速度、延迟和上下文窗口等性能指标 [1]。其中开源模型占94个 [1]。
排行前五分别为：Claude Opus 5（61分）、Claude Opus 5 Xhigh Effort（60分）、Claude Fable 5（60分）、GPT-5.6 Sol（59分）和Claude Opus 5 High Effort（59分） [1]。在开源模型中，GLM-5.2 (max)排名最高，得分51分 [1]。
在其他性能指标方面，Mercury 2实现了最快的输出速度，达901.6 tokens/秒 [1]；Nova Micro提供了最低价格，为每百万tokens 0.03美元 [1]；Gemini 2.5 Flash-Lite的首token延迟最低，为0.33秒 [1]。
Claude Opus 5 with Adaptive Reasoning and Max Effort mode has claimed the top position on the Artificial Analysis Intelligence Leaderboard, achieving a score of 61 points [1]. The leaderboard conducted a multidimensional comparative assessment of 586 models across various performance metrics, including intelligence level, pricing, output speed, latency, and context window capabilities, with scores derived from evaluations of 170 models [1].
The top five performers on the leaderboard are Claude Opus 5 with 61 points, Claude Opus 5 Xhigh Effort with 60 points, Claude Fable 5 with 60 points, GPT-5.6 Sol with 59 points, and Claude Opus 5 High Effort with 59 points [1]. Among the 170 models evaluated, 94 are open-source [1].
Beyond overall intelligence rankings, the leaderboard tracks performance across specialized categories: Mercury 2 leads in output speed at 901.6 tokens per second, Nova Micro offers the lowest pricing at $0.03 per 1 million tokens, and Gemini 2.5 Flash-Lite achieves the lowest first-token latency at 0.33 seconds [1]. Among open-source models, GLM-5.2 (max) ranks highest with a score of 51 points [1].