Fine-tuned open-source models cost 95% less and beat frontier models, claims dev

ayushtweetshere · x · 2026-09-13

A viral thread claims fine-tuned open-source models on custom data cost 95% less than frontier models, perform better, and can be trained in under 48 hours — arguing frontier labs have no moat even in enterprise, and will resort to FUD (fear, uncertainty, doubt) as open source scales. The retweeter adds that the future may not even be LLMs and open source is the only fix, plus unverified speculation about DARPA influence. Treat the performance claims as unverified single-source opinion.

Original post →

Fine-tuned open-source models cost 95% less and beat frontier models, claims dev

More from Venture