elvis · @omarsar0 · X·2026-09-16 02:47·9小时前
AI 导读

TypeSafe 发布新架构模型 Jev,放弃文本生成,通过并行回答结构化问题,并用 RLCD 训练其概率以反映正确频率,在 System One 查询上响应速度提升 40-200 倍。据引用推文,Jev 号称快 20-200 倍、便宜 40-400 倍(输出 token 免费),定位为面向决策的前沿可组合智能。

elvis@omarsar0
36AI 编辑部评分,满分 100
2026-09-16 02:47· 9小时前
AI 导读

TypeSafe 发布新架构模型 Jev,放弃文本生成,通过并行回答结构化问题,并用 RLCD 训练其概率以反映正确频率,在 System One 查询上响应速度提升 40-200 倍。据引用推文,Jev 号称快 20-200 倍、便宜 40-400 倍(输出 token 免费),定位为面向决策的前沿可组合智能。

Recommended read. Jev gives up text generation to make AI dramatically faster.

TypeSafe built a new architecture that answers structured questions in parallel, with RLCD training its probabilities to reflect how often it’s right.

The team reports 40–200x faster responses on System One queries.

Diogo AlmeidaAfter co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to t...

来源:elvis· x.com