Skip to content
D-CSIL

Model Registry · HELIX family

HELIX-120M

H-E-R stack · 15 layers · d_model 768 · 12 heads · MoE top-2 of 4

STATUS PLANNEDPROJECT: HELIX
HELIX-120M — TRAINING LOSS (DEMO DATA)1 STEPS · MIN 10.90
10.910.910.910.910.90k0k0k0k0k

Benchmarks

MetricValueNotes
Memory budget (design)≤13.3GBsafety ceiling verified in config

Known Limitations

  • Not yet trained — capacity-control variant awaiting the 200M run.