AMAZINGINDEX.COM 每天向上一点
53.9
VOL. 2026.11
2026.11.10
← 返回 2026.11.10 日报
日报快照 · Daily Snapshot
NO. 020

Claude Fable 5.2登顶智能榜但性价比垫底

#ARTICLE HackerNews 2026.11.10
推荐指数 49.0 NO. 020 · 2026.11.10
发布2026/11/09Score271Comments190

Claude Fable 5.2在Artificial Analysis智能指数拿下65分,超越所有竞品,但定价$8/百万token是行业均价的4倍,且生成token量高达中位数的3.8倍。高智能伴随高成本和高延迟,企业需要权衡是否值得为边际智能提升买单。

这个评分结果暴露了当前大模型市场的核心矛盾:智能天花板和推理成本之间的剪刀差在扩大。Anthropic显然选择了用极端的verbosity换准确率,275M token的生成量意味着实际调用成本可能接近$20-30/次复杂任务,这对需要高频调用的生产环境是致命的。

更深层的问题是,当OpenAI、Google都在推轻量高效模型(GPT-4o mini、Gemini Flash)时,Anthropic反其道而行之的底气在哪?可能是瞄准了金融、法律等容错率极低、愿付溢价的垂直场景。做模型选型的团队要注意,这个模型不适合做客服、内容生成等高并发场景,但在合同审查、合规分析等低频高价值任务上可能值得测试。

意见分歧 161 条评论

核心争论:Max设置性价比低是否代表模型本身不值,还是用户应主动选择中低设置

qntty

Crazy how quick these are coming out now

jrflo

I'm sure that they get early access to run the benchmarks, it wouldn't have time to finish running if they were running it at public release

AnotherGoodName

Someone on a thread was complaining they haven’t really needed the newer models because Opus 4.6 was good enough. I had to remind them 4.6 only came out February this year. In a normal world 4.6 is a newer model! A model 6 months old is seen as ancient!

替代方案: Opus 4.6Sol 6.1ArgonAragonCarbon
查看原文 →