Anthropic于2026年11月9日发布了Claude Fable 5.2模型 1。根据Artificial Analysis发布的最新数据,该模型在Max模式下的智能指数得分为65分,创下该榜单的历史新高 1。这一成绩显著高于同类模型41分的中位数水平 1。尽管性能卓越,但该模型相比同价位竞品存在明显短板,其单次评估平均耗时为45.3分钟,远高于中位数的13.1分钟 1。此外,该模型的推理速度较慢且输出冗长 1。
在API定价方面,Claude Fable 5.2的成本高昂,输入价格为每百万tokens 8.00美元,输出价格为每百万tokens 40.00美元 1。这两项价格均大幅超出中位数水平(分别为2.00美元和10.00美元) 1。该模型支持文本和图像输入,拥有100万tokens的上下文窗口 1。报告详细列出了其在不同努力级别下的表现:Max、Xhigh、High、Medium和Low五个级别的得分分别为65、62、62、56和54 1。
Anthropic has released its new large language model, Claude Fable 5.2, which achieved a score of 65 on the Artificial Analysis Intelligence Index in Max mode, marking the highest result recorded for that benchmark to date 1. The model was launched on November 9, 2026, and significantly outperforms the median score of 41 among comparable models 1. Scores varied by effort level: Xhigh and High both scored 62, Medium scored 56, and Low scored 54 1.
Despite its superior intelligence metrics, the report highlights substantial trade-offs regarding cost and latency. Claude Fable 5.2 is priced at $8.00 per million input tokens and $40.00 per million output tokens, rates that are considerably higher than the market medians of $2.00 and $10.00, respectively 1. Additionally, the model exhibits slower inference speeds, with an average evaluation time of 45.3 minutes per run, compared to the median of 13.1 minutes 1. The system supports both text and image inputs and features a context window of one million tokens 1.
评论
还没有评论,欢迎留下第一条。