Layer1Labs is raising seed funding — $1–1.5M on a SAFE, seeking a lead investor. Talk to us →
Document · ChronoHive

ChronoHive Measured Results

What ChronoHive has measured, how each figure was obtained, and the limits published alongside it — every figure printed with the basis it was measured on: the workload, the hardware, and the regime.

Version 1.0 · October 26, 2026 · PDF · 8 pages · Layer1Labs Silicon Inc.

Key figures

+9.8%

end-to-end training speedup on real data — just under the trial’s own 10% confirmation bar, reported exactly as measured

Measured — preregistered A/B

+24.1%

wall-clock improvement in the slow-storage regime (RTX 5060 Ti; +21.7% on the RTX 4070 SUPER)

Measured

92 of 92

runs reproduced exactly, and 400 of 400 checkpoints verified

Measured — determinism

5–15%

the speedup range the model projects across storage regimes

Modeled — not a measurement

What it covers

  1. How to read these results
  2. The headline result
  3. Results by storage regime
  4. Determinism, integrity, and checkpointing
  5. Admission control, and a contention demonstration
  6. What is modeled, not measured
  7. Limits, stated plainly
  8. What these results do not establish

Basis

This document collects the results published on the ChronoHive page, in one place, with the labels the measurements carry on the website. Gains are never stacked across regimes, and every figure travels with its basis.

The misses are here too: the regimes where ChronoHive currently trails the conventional approach are stated in the document, in the same terms the website’s FAQ states them.