BENCHMARKS

Evidence-bounded benchmark reporting.

ZETAPHI benchmark statements are scoped to custom models, specific test regimes, and matched comparisons. This page is designed to expand as more public benchmark material is finalized.

CURRENT POSITIONING

What we can say cleanly right now

  • Work is focused on attention-replacement architecture directions for custom models.
  • Primary emphasis is linear scaling and compute-speed advantages in relevant settings.
  • Capability claims remain benchmark-dependent and evidence-scoped.

CLAIM STANDARD

Matched comparisons only

Public performance language should be interpreted in benchmark-specific context, not as a universal claim across all model families or tasks.

NEXT ADDITIONS

Public-facing benchmarks will be posted soon