Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding
Article URL: https://www.ricursive.com/ripple Comments URL: https://news.ycombinator.com/item?id=50369925 Points: 272 # Comments: 99
Ricursive Intelligence unveiled Ripple-65B Preview, a cutting-edge language model achieving U.S. open-weight state-of-the-art (SOTA) on agentic coding. This full-stack AI lab, specializing in self-improving systems through hardware, software, and model co-optimization, trained Ripple-65B Preview for a mere $32 million, one-tenth the estimated cost of the previous open-weight SOTA.
The model was trained on 12 trillion tokens with a capacity for up to 1 million context length, delivering impressive results on agentic coding benchmarks. With an 80.52 score on TerminalBench 2.1, Ripple-65B Preview outperforms larger models such as Reflection Beam 501B, Meta Muse Spark Open 816B, Thinking Machines Inkling 975B, and NVIDIA Nemotron 3 Ultra 550B.
Its dense architecture provides strong reasoning performance with low-latency token generation, enabling efficient execution on local devices like NVIDIA DGX Spark. Ripple-65B Preview is accessible on Hugging Face, Together AI, GMI Cloud, and OpenRouter, while Fireworks, its pretraining pipeline, goes live later.
Written by urgent.news from Hacker News Best's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
