Small Models and Quantization

[plitc]

Hello MiniMax Team,

will you also be offering smaller model variants in the future?
Ones that fit into 200 GB or 100 GB VRAM setups and still deliver acceptable throughput (t/s)?

best regards

[dysangel]

support for Apple Silicon would also be fantastic. I have a 512GB Mac Studio and would love to run a smart model with linear attention requirements