用户实测 Muse 跑复杂工程任务胜过 Kimi K3、GLM 5.3
一位用户反驳「AI 只会做游戏」的说法，称自己用复杂工程测试对比顶级模型，Muse 表现优于 Kimi K3 和 GLM 5.3，并指出 Muse 与 Gemini Flash 一样是为自家 harness 训练的。
Bro thinks ai only makes games 😭 muse is a good model for alot of things. You wouldn't know cuz you dont actually code. U hust sit on stream and tell ur agents to check bridge mind bs every single day while clipfarming by posting controversial titles that are exaggerated. In many ai chats everyone points out how lame this behavior is. Because u have meta working super hard on a model. Its cheap and good. Sure it fails a lava lamp test, btw u never posted prompt or harness. I recently ran an actual complex engineering test on all the top models including muse. And muse did a really good job & beat many other top models like kimi k3 and glm 5.3 Also most of these models, like Muse and Gemini Flash, are very much trained for their native harnesses. Same thing with DeepSeek but you never mention these things because they don't get you clicks As a popular account for seeing ai, shifting on every model all the time for clicks is just so lame bro. When bijan Bowen shows a model being ass. He gives it a hones go. Also, the benchmarks are for muse 1.3 MAX u prolly arent even using max if ur not using muse code with muse sub 😭