#data-efficiency

A March 19, 2026 Hacker News post about NanoGPT Slowrun reached 162 points and 43 comments at crawl time. Q Labs says an ensemble of 1.8B-parameter models trained on 100M tokens matched a baseline that would normally require 1B tokens.

A March 4, 2026 Hacker News thread elevated Q Labs’ Slowrun benchmark, which fixes training data at 100M FineWeb tokens and optimizes for data efficiency under large compute budgets.