gemini-4POST

谷歌内部对 Gemini 4 编码与真实任务表现存疑

消息称尽管 benchmark 成绩亮眼，谷歌员工仍质疑 Gemini 4 在编码等关键任务上的表现，认为其真实场景落地效果偏弱。

First Squawk

@FirstSquawk28 天前收录原推发布于 2026.10.01

First Squawk @FirstSquawk原推发布于 2026.10.01

GOOGLE IS GRAPPLING WITH INTERNAL SKEPTICISM OVER GEMINI 4 PERFORMANCE, WITH EMPLOYEES RAISING CONCERNS ABOUT THE MODEL'S PERFORMANCE IN CODING AND OTHER KEY TASKS DESPITE STRONG BENCHMARK RESULTS, AND GEMINI 4 REPORTEDLY PERFORMING LESS WELL WHEN PUT TO WORK ON REAL-WORLD TASKS.

x.com/firstsquawk/status/2105387845930324058采集于 2026.10.01 10:02