Reaction
Robot-Use Agents with DeepSeek V4.1 Flash draw praise and a model-choice caveat
One viewer says the results exceeded expectations; another praises them but calls V4.1 far from the best vision-language model.
TLDR
One viewer says Robot-Use Agents using DeepSeek V4.1 Flash delivered results beyond their expectations. Another calls the showing “really cool” and argues that V4.1 is economically the best Chinese model, but not remotely the best vision-language model. They also wonder whether Hangzhou will have an edge in robotic training for scenarios hard to model in MuJoCo.
Combined views
1.5K
3 Sources, first seen ago
