museVIDEO

Meta Muse Wonder retested with zebra prompt, falls short of GPT-6.1 Sol

The author swapped the possibly over-trained pelican prompt for a zebra riding a bicycle test: Meta Muse Wonder produced an image in 42.1 seconds at 201 tps but with poor results, while GPT-6.1 Sol took 22 minutes 33 seconds at 36 tps and was more accurate.

雪瑜

@xueyu1125Archived 22 days agoOriginally posted 2026.10.13

雪瑜 @xueyu1125Originally posted 2026.10.13

Open original on X

第二波Meta Muse Wonder实测来了，Meta Muse Wonder好像原形毕露了，绷不住了朋友让我换个提示词，鹈鹕测试现在可能被模型专门训练过了，我改了个斑马骑自行车做测试 - Meta Muse Wonder 201tps 用时42.1秒 - GPT-6.1 Sol 36tps 用时22分33秒 GPT可以看到是在说太慢了，但是整体效果确实要好一些，Muse Wonder拉完了🤣

This post quotes another post: “Meta Muse Wonder vs GPT-6.1 Sol: motion strong, aesthetics behind”

x.com/xueyu1125/status/2110019169496568136Archived 2026.10.14 10:30