STATE: IDLE
LAYER: —
TOKENS: 0
ATTENTION: 0 heads
Model Scale
ModelGPT-4
Context Window128K
Layers96
Attention Heads96 / layer
Parameters175B
Self-Attention Cost
Attention Matrix16.4B cells
Memory (KV cache)~98 GB
Key Differences
⚡ No spontaneous activity
→ Feedforward only (no loops)
🧊 Weights are frozen after training
📦 Processes entire input at once
🔢 Everything is matrix multiplication
Generated Output
Type a prompt to begin...
Gemini API Key
SIMULATED
Stored locally · Never sent anywhere except Google's API
Press Enter to send · Watch the signal flow through 12 transformer layers