RicursiveAI 发布从零预训练的 65B dense 模型 Ripple-65B Preview，称其在美国开源权重模型的 agentic coding 任务上达到 SOTA，超过 Muse Spark Open-816B、Beam-501B、Inkling-975B 和 Nemotron 3 Ultra-550B 等更大模型。swyx 引用称其完整研发成本仅 3200 万美元，为上一开源权重 SOTA 的十分之一，涵盖预训练阶梯开发、数据配比、评测、合成数据生成与后训练配方等环节。官方博客称通过模型阶梯实验把代码任务的算力效率提升至多 16 倍。
Ripple-65B Preview 以 3200 万美元的全流程研发成本，在 agentic coding 上超过多个更大规模的开源权重模型。
we developed and trained Ripple-65B Preview for only $32M, which is one-tenth the cost of the previous open-weight SOTA on agentic coding¹
this is like the avengers of open model SOTA
¹ "Note that $32M is our full R&D cost starting from scratch, including pretraining ladder development, data mixing, evals, debugging, synthetic data generation, posttraining recipe development, and the final hero pretraining and posttraining run."
Today we release Ripple-65B Preview, our 65B dense LLM we pretrained from scratch. Ripple sets a new state-of-the-art for U.S. open-weight models in agentic coding tasks, surpassing significantly larger models including Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B. Learn more: https://www.ricursive.com/ripple在 X 查看被引用的帖子
来源：swyx · x.com