Event ·Multi-source

TypeSafe AI 发布 Jev:只输出浮点数的决策模型

TypeSafe AI 上周发布 Jev,称其属于新类别“System One 模型”,也可理解为决策模型。Jev 接收文本输入,但输出的是对应类别、是/否判断、评分及置信度的浮点数而非文本。定价上 Jev 只收输入费、输出免费,首个模型输入价为每百万 token $0.042,低于 OpenAI GPT-5 Nano 的 $0.05。API 支持三类问题:基于 Bernoulli 分布的 Noul 是/否题、多选项概率分布题、数值评分题,多个问题并行评估。作者 Simon Willison 用它做搜索重排——先用 BM25 取回 100 条候选再让 Jev 打相关性分,同时指出这类模型比 LLM 更接近黑盒。

Current read

TypeSafe AI 的 Jev 不输出文字只输出浮点数,输入每百万 token 才 $0.042,做分类和搜索重排能省不少钱。

11 sources53° AI heatLast updated 2026/9/21 23:09:20

Evidence

11 sources
Primary sourceSimon Willison’s Weblog
View source
Related source 1掘金本周最热
View source
Related source 2shao__meng
View source
Related source 3OpenRouter
View source
Related source 5歸藏(guizang.ai)
View source
Related source 8Sam Altman
View source
Related source 9lmarena.ai
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.