for vision multimodal chat conversational model in 29m you may have a look on minimind

[cyysky]

i think can try to made this model in english with MiniMind2-Small-V architecture and modified it to working in ESP32😃

https://github.com/jingyaogong/minimind-v

[huzongyao]

could do
https://www.bilibili.com/video/BV11rGg6MEVx