This new MoE AI model looks promising with 1T parameters (32B activated)
The results speaks for themselves.