Open Source Models Write Better Than OpenAI/Anthropic, User Claims

JoshPurtell · x · 2026-08-20

User JoshPurtell argues that models from major labs are "atrocious" at writing. He observes that open-source models like Nemotron, which likely lack extensive writing-specific RLHF beyond DPO, write significantly better than proprietary models, questioning whether labs actually train for writing quality.

Original post →

More from Models

Models channel →