{"version":1,"type":"story","url":"https://digestai.news/story/ricursive-intelligence-releases-ripple-65b-preview-a-65b-model-with-1m","json":"https://digestai.news/story/ricursive-intelligence-releases-ripple-65b-preview-a-65b-model-with-1m.json","markdown":"https://digestai.news/story/ricursive-intelligence-releases-ripple-65b-preview-a-65b-model-with-1m.md","slug":"ricursive-intelligence-releases-ripple-65b-preview-a-65b-model-with-1m","headline":"Ricursive Intelligence releases Ripple-65B Preview, a 65B model with 1M context","summary":"Ricursive Intelligence announced Ripple-65B Preview, a dense Transformer with 65B parameters trained on 12 trillion tokens and supporting up to 1M context length. The model is offered in BF16 and NVFP4 numerics, with an instruction‑tuned version and a base BF16 checkpoint.\n\nIn agentic coding benchmarks the model scored 80.52 on TerminalBench 2.1, outpacing larger open‑weight models such as Reflection Beam 501B (80.1) and Meta Muse Spark Open 816B (76.4). Compute efficiency on code improved 16x versus the Olmo 3 ladder and rises to 55x for the full run, while maintaining 40% MFU throughout training. The architecture uses a narrow 5120 hidden dimension across 80 layers and runs at about 250 tok/s on 2xB200 GPUs with batch=8. The team grew from 1 to 5 people and the model can fit on consumer hardware like an NVIDIA DGX Spark, with a public NVFP4 checkpoint and colab to help users start.","keyPoints":["Ripple-65B Preview has 65B parameters, 12 trillion pretraining tokens, and 1M context length","Scores 80.52 on TerminalBench 2.1, beating larger US open‑weight models","Training efficiency improves 16x vs Olmo 3 ladder, reaching 55x at full scale"],"whyItMatters":"A 65B open‑weight model with 1M context offers strong coding performance while remaining runnable on consumer‑grade hardware, expanding access to high‑capacity LLMs.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Ricursive Intelligence"],"models":["Ripple-65B Preview"],"people":[]},"firstPublishedAt":"2026-11-04T19:16:16Z","updatedAt":"2026-11-04T19:16:16Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"huggingface.co","title":"ricursiveAI/Ripple-65B-Preview · Hugging Face","url":"https://huggingface.co/ricursiveAI/Ripple-65B-Preview","publishedAt":"2026-11-04T19:16:16Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1xnx3cs/ricursiveairipple65bpreview_hugging_face/","points":null}],"thread":null,"cite":{"text":"Digest AI, \"Ricursive Intelligence releases Ripple-65B Preview, a 65B model with 1M context\", 4 November 2026, https://digestai.news/story/ricursive-intelligence-releases-ripple-65b-preview-a-65b-model-with-1m","publisher":"Digest AI","title":"Ricursive Intelligence releases Ripple-65B Preview, a 65B model with 1M context","datePublished":"2026-11-04T19:16:16Z","url":"https://digestai.news/story/ricursive-intelligence-releases-ripple-65b-preview-a-65b-model-with-1m"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}