Get the latest tech news

Training a 3.8B LLM to 0.384 CORE for $998 – Hugo Vergnes


A 3.8B model pretrained from scratch to 0.384 CORE on 65B tokens in 43 hours, for $998.

None

Get the Android app

Or read this on Hacker News

Read more on:

Photo of Core

Core

Photo of 3.8B LLM

3.8B LLM

Photo of Hugo Vergnes

Hugo Vergnes

Related news:

News photo

'Creative magic use was at the core of what we wanted to do in gameplay, and we landed on warlocks' — Invoke Studios head explains why it chose to highlight the arcane class in Warlock: Dungeons & Dragons

News photo

AMD reportedly prepping Ryzen 5 7500 (non-F) CPU with integrated graphics at double the price — Six-core Zen 4 chip rumored to share identical specs with its F-moniker cousin

News photo

Intel's Core Ultra 400 'Nova Lake' launch schedule leaks out — mass production in Q4, first Nova Lake CPUs in Q1 2027