NanoGPT Speedrun Frontier (human record gap closed at 24 agent-hours)
Agents · 2026-08-15
Prime Intellect's NanoGPT Speedrun Frontier runs a model unattended in an agent harness for up to about nine days on the modded-nanogpt track 3 optimizer speedrun. On one 8xH200 node whose only network route is the model's API, the agent changes the training recipe of a 124M-parameter GPT to reach validation loss 3.28 in as few training steps as possible, and a frozen verifier accepts a record only if the mean loss of eight trainings on fixed seeds beats 3.27859. This component scores the same share of the 690-step gap from the best validated record that each model's best final run had reached after 24 agent-hours, the board's equal-budget comparison; runs that ended sooner have no value. It is a different measurement from OpenAI's internal NanoGPT self-improvement figure (nanogpt-openai-reported), which uses track 1 on a single H100 for under a day.
Benchmark source: Primeintellect