Event ·Multi-source

Ricursive Intelligence 发布 Ripple-65B Preview 开放权重编程模型

前谷歌研究员创立的 Ricursive Intelligence 于 11 月 4 日发布 Ripple-65B Preview，宣称在美国开放权重模型中智能体编程表现最佳。该模型在 TerminalBench 2.1 上超过 Meta Muse Spark Open 816B 和 NVIDIA Nemotron 3 Ultra 550B 等更大规模对手。Ripple-65B Preview 是稠密模型，在 12T Token 上预训练，支持 1M 上下文长度，从头开发训练成本约 3,200 万美元。该模型此前版本专用于 AI 驱动的芯片设计平台。

Current read

一个 65B 的小模型在 TerminalBench 上打赢 816B 的大块头，还支持 1M 上下文，写代码的可以试试权重。

11 sources46° AI heatLast updated 2026/11/6 02:51:50

Evidence

11 sources

Primary sourceIT之家

2026/11/6 02:51:50View source

Related source 1elvis

2026/11/4 22:18:15View source

Related source 2Stanford AI Lab

2026/11/4 19:26:40View source

Related source 3NVIDIA AI

2026/11/4 18:56:55View source

Related source 4vLLM

2026/11/5 14:14:59View source

Related source 5rohanpaul_ai

2026/11/5 20:48:25View source

Related source 6@atomic_chat_hq

2026/11/4 17:53:37View source

Related source 7arXiv cs.AI

2026/11/5 03:41:53View source

Related source 8techcrunch

2026/11/5 14:00:00View source

Related source 9Blognone

2026/11/5 23:28:54View source

Related source 10AIGCLINK

2026/11/4 02:54:56View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story