Not anywhere close to Opus in terms of breadth of skills

So I tried the new 4.1 and I am disappointed. It is possible that it's just benchmaxxed. It is good at skills it's been trained to do, but it has not been trained in as many areas as Opus 5.5 is. Or maybe Opus having a few trillion parameters vs Flash's sub trillion parameters makes Opus superior. Anyway, the new DeepSeek is better at what it could already do, but it's still like its previous version at things it was struggling to do (e.g. reverse engineering old games and porting them to Rust). Just my 2 cents.

添加评论
点赞收藏
点踩分享查看原文
评论
?
参与讨论