Key Info
GMI Cloud revisited a May comparison between DeepSeek V4 Pro and Xiaomi MIMO V2.5 Pro, noting that open-weight models have advanced dramatically in just five months when given the same prompt.
Highlights
- In the May test, MIMO V2.5-Pro beat DeepSeek-V4-Pro by finishing about 1 minute faster and producing a cleaner aesthetic.
- The team attributed MIMO's edge to native multimodality, giving it a stronger sense of good visuals.
- Re-running the same prompt five months later, open-weight models went from beginners to pros at gaming and 3D world simulation.
- The author wonders how DeepSeek V4 Vision might change the comparison.