# Decentralized training of frontier models is not possible today

- **Speaker:** Greg Osuri
- **Attribution:** attributed
- **Said:** 2025-04-18
- **Theme:** Decentralized AI — https://www.gregosuri.com/predictions/decentralized-ai/
- **Source:** DePin, Scams & Decentralized ML (Chris Joannou)
- **Watch at:** 00:43:05 — https://www.youtube.com/watch?v=ImwkqmOSUkE&t=2585s
- **Canonical:** https://www.gregosuri.com/predictions/decentralized-ai/2025-04-18-decentralized-training-of-frontier-models-is-not-p-bglh/
- **Cite as:** Greg Osuri, "Decentralized training of frontier models is not possible today," DePin, Scams & Decentralized ML (Chris Joannou), 18 Apr 2025, 00:43:05. https://www.youtube.com/watch?v=ImwkqmOSUkE&t=2585s

## Quote — verbatim

> If you want to train advanced models, you need low latency, very minimal message-passing latency between two neurons. So if you have to do it in a decentralized way, that's not possible... the trade-off is, yeah, you can get lower cost, but training time is going to be higher. So you're never going to catch up to a GPT-4.

## Context — site annotation, not spoken

Honest technical framing: communication cost must be lower than compute cost (00:34:30); fine-tuning and inference distribute fine, frontier training doesn't — yet. He praises Gensyn for attacking this.