AI Foundationspredict · compress · act

concepts → Model

Model

DeepSeek-V3

frontieras of 2026-094 connections

a 671B MoE (37B active) trained in FP8 — frontier-adjacent at a fraction of the assumed cost

Where it sits

The prism has six jobs across and eight layers down. Its primary cell isno mode × L1, with a noted second cell. Hatched cells cannot exist — a GPU does not learn, an institution does not infer.

—×L7 — it is also an ecosystem event: the 2025 open-weights turn

What must come first — and what it unlocks

Left to right is reading order, derived from the prerequisite_of edges. Nothing here is hand-ordered: the diagram is the graph.

Mixture of expertsDeepSeek-V3DeepSeek-R1

Before it: Mixture of experts

Unlocks: DeepSeek-R1

Where to read it

The chapter that introduces it, and any chapter that uses it again.

45DeepSeekact 10 · The Open-Weight World

Where it comes from

paperDeepSeek-V3 Technical ReportDeepSeek-AI · 2024

Every connection

All 4 edges touching this node, grouped by relation family — the sections above are highlights from this list. Colours match the relation families inthe atlas.

Order · 2
requiresMixture of expertsMechanism
must come beforeDeepSeek-R1Model
Flow · 1
usesMixture of expertsMechanism
Lineage · 1
is improved byDeepSeek-V4Model

This page is a projection of one node in src/data/concepts.ts. It has no prose file of its own — 230 declared edges produce all 236 of these pages. Edit an edge and both endpoints change.