← Back to blog
Demo 2 / 3
Sparse Activation Explorer
Configuration
Total experts (N)
64
Active per token (k)
2
Quick presets
Switch-T
8 exp, top-1
Mixtral
64 exp, top-2
Gemini MoE
160 exp, top-6
Large MoE
256 exp, top-8
Live metrics
Expert pool — active experts highlighted per token
Grid view:
All experts
Sample (64)
▶ Animate token stream
Click any cell to see its index
Active
Idle
Token routing simulation — each token activates a fresh set of top-k experts
Incoming tokens
Click "Animate token stream" to route tokens
Compute & memory comparison vs. equivalent dense model