In tests announced by the team, OpenBMB’s newly released MiniCPM-V 4.7 35B-A3B version can generate more than 13 tokens per second on an iPhone 17 Pro Max, using about 3GB of memory while running.
That 3GB refers to runtime memory; the model files require additional storage space. It preloads model weights that may be needed next, reducing the time spent waiting for data.
That 3GB refers to runtime memory; the model files require additional storage space. It preloads model weights that may be needed next, reducing the time spent waiting for data.