Geek (@geekbb) “我们讲究的是一个花小钱办大事，其实还是挺赞成他的说法。 把 V4 Pro 的落差理解成欠” — TopicDigg
Geek
@geekbb
加入 July 2010
0 正在关注 0 粉丝
我们讲究的是一个花小钱办大事，其实还是挺赞成他的说法。
把 V4 Pro 的落差理解成欠训，也比“架构不行”更贴公开数字。
V4 已经没有 R1 那种突然改写叙事的空间，大家默认它该贴美国一线，结果只是“接近、更便宜、上下文更长”，体感就会写成不够惊艳。
算力短缺是一个说得通的主因，但还混着欠训、后训练和预期过高。
显示更多
Deepseek-V4-Pro-0813 doesn't seem as insanely impressive as I expected.
This seems to be a pattern with Chinese AI labs:
smaller models like Qwen 27b, Deepseek-V4-Flash, and GLM-5.2 perform ridiculously well for their size,
but maybe due to a lack of training compute, their larger models feel a bit unaligned.
As they secure more compute, they will definitely get better over time.
显示更多
