Ripple-65B Preview发布：Apache 2.0开源美国模型，智能体编码表现强劲
原标题：Ripple-65B Preview delivers strong agentic coding results as a fully Apache 2.0 open US model
AI 摘要
AI芯片设计公司Ricursive Intelligence训练并发布了开源语言模型Ripple-65B Preview，该模型基于12万亿token预训练，采用65B参数稠密Transformer架构，以Apache 2.0许可证在Hugging Face上开放。其在TerminalBench 2.1编码基准上得分80.52，为美国开源模型最佳，但SWE-Bench Pro v2.0 Hard、DeepSWE v1.1和MCP Atlas等多项成绩落后于同期发布的Inkling 2 Small。公司称全部研发与训练成本仅3200万美元，模型针对Nvidia硬件优化，更易于在用户自有硬件上部署。
正文节选
Ripple-65B Preview delivers strong agentic coding results as a fully Apache 2.0 open US model AI chip design firm Ricursive Intelligence has trained an open language model. Ripple-65B Preview was pretrained on 12 trillion tokens and, according to the company, achieved a score of 80.52 on the TerminalBench 2.1 coding benchmark, the best result among open US models. Ripple partly uses Terminus 2 and partly OpenCode as its harness. Other benchmark results show that Ripple does not match the also ne