Event ·Multi-source

Recursive 发布 Flint:本地编码模型,MacBook Pro 上跑出 125+ tokens/sec

Recursive SI 发布首个模型 Flint,定位本地快速编码模型,由其自动化研究系统 Eureka Machine 构建。官方称速度最高可达同类模型的 2 倍:MacBook Pro 上超过 125 tokens/sec,RTX Spark 上超过 150 tokens/sec。模型从零开始训练,用时不到 4 个月,权重已在 Hugging Face 开放,可在本地框架运行,也可在 OpenCode 或 Pidot 上免费试用。

Current read

Richard Socher 团队出了个本地编码模型 Flint,MacBook 上每秒能跑 125+ tokens,权重开放,不用联网不用按量付费,写代码的可以试试。

7 sources47° AI heatLast updated 2026/10/21 16:29:32

Evidence

7 sources
Primary sourceRichard Socher
View source
Related source 1TestingCatalog
View source
Related source 2rohanpaul_ai
View source
Related source 3marktechpost
View source
Related source 4techcrunch
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.