DeepSeek-V3.2 — full network dataflow (official inference reference)

Built from deepseek-ai/DeepSeek-V3.2 inference/model.py (ground truth), config.json, and arXiv:2512.02556 (Gemini 3.1 Pro extraction, adversarially verified). Blue = GEMM (precision noted), teal = elementwise/norm, red = attention core, orange = caches, purple = routing/index ops. Drag to orbit · scroll to zoom · press ? for keys.
Key dimensions — DeepSeek-V3.2
hidden H = 7168 · 61 layers (3 dense + 58 MoE) · vocab 129280
MLA: 128 heads · q_lora 1536 · kv_lora 512 · nope 128 + rope 64 · v 128
KV/token: 512 c_kv (FP8) + 64 k_pe — MQA-absorbed decode
indexer: 64 heads × 128 · Σ w·ReLU(q·k) in FP8 · top-2048
MoE: 256 experts, top-8 via 4/8 groups · inter 2048 · shared ×1 · ×2.5
FP8 e4m3 block 128×128 (ue8m0) · YaRN 163840 · MTP ×1 (paper; absent in ref code)
FC / GEMM (precision as labeled)
elementwise / norm / reshape
attention / score core
KV cache / persistent state
routing / gating / speculative ops
embeddings / logits / block refs
─ solid: dataflow · ┄ dashed: control/index · ┄ orange: cache update
✕ close
keys: 15 boards · ? all shortcuts · hover / click nodes
Keyboard controls
15 fly to board  ·  0/H overview  ·  [ ] prev / next board
or WASD pan  ·  Q/E or -/= zoom
hover node → details  ·  click node → focus camera on it
T transcript  ·  F flow pulses  ·  L bandwidth heatmap  ·  B bloom  ·  V trails  ·  K depth-of-field
M mouse mode (orbit ⇄ pan-first)  ·  I ASCII render mode  ·  G guided tour  ·  U pin transcript in 3D
P PNG  ·  ⇧P PNG 2× supersampled  ·  O SVG vector  ·  X .glb scene
? this help  ·  Esc close overlays