NYX 
EDGE INFERENCE
NyxCore
Language models built into FPGA logic, answering from grounded sources where there is no cloud and no link.
1.3 BPARAMETERS
1CARD
TernaryWEIGHTS
U50AMD ALVEO
What it does
Ternary neural networks are implemented directly in FPGA logic. Each weight takes one of three values, so the datapath adds, subtracts or skips where a general-purpose accelerator would multiply. That cuts power and memory traffic and keeps timing fixed.
A grounded retrieval harness with constrained decoding keeps the model answering from its sources, so it cannot answer outside them.
The result runs where there is no cloud and no link, on a single card.
Where it runs: AMD Alveo U50.
Evidence
1.3 B
Language model inference on adaptive silicon measured
A ternary model running on a single card with grounded retrieval and four deployed knowledge bases, unable to answer outside its sources.
AMD Alveo U50