Understand this story
On November 5, 2026, Nathan Lambert commented on Ricursive Intelligence's release of Ripple-65B Preview, calling it an ambitious project and expressing hope that more U.S. labs would follow suit, building solid models with small, focused teams. He noted that the model's approach continues the olmo series, and that seeing a project he was involved in scaled up is very meaningful. On the same day, Ricursive Intelligence released this 65B dense large model pretrained from scratch, claiming it achieves SOTA on agentic coding tasks among U.S. open-weight models, surpassing larger models such as Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B. swyx quoted its explanation stating that the $32 million is the full R&D cost from scratch, covering pretraining ladder development, data mixing, evaluation, debugging, synthetic data generation, post-training scheme development, and the final pretraining and post-training runs.
AI-generated from public reports
Latest development · Report
Ricursive Intelligence Releases Ripple-65B Preview, Training Cost $32 MillionDevelopments
2 confirmed developments
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 5
Swyx
Ricursive Intelligence Releases Ripple-65B Preview, Training Cost $32 Million
Ricursive Intelligence has released Ripple-65B Preview, a 65B dense large model pretrained from scratch, claiming it achieves SOTA on agentic coding tasks among U.S. open-weight models, surpassing larger models such as Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B. swyx cited its explanation stating that the $32 million is the complete R&D cost from scratch, covering pretraining ladder development, data mixing, evaluation, debugging, synthetic data generation, post-training scheme development, and the final pretraining and post-training runs.
@mark_kRelated discussion
@mark_k joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Hacker News Trending
Ricursive Intelligence Releases Ripple-65B Preview Open-Weight Model
Ricursive Intelligence has released Ripple-65B Preview, a general-purpose open-weight model that scores 80.52 on TerminalBench 2.1, which it claims is SOTA among US open-weight models, surpassing Reflection Beam 501B's 80.1 and Meta Muse Spark Open 816B's 76.4. The model was pretrained on 12 trillion tokens, supports a context length of up to 1M, uses a dense Transformer architecture rather than MoE, and had a total R&D cost of $32 million. The model is now available on Hugging Face, Together AI, GMI Cloud, and OpenRouter, with Fireworks launching later the same day.
Nathan LambertSelected
Nathan Lambert on Ripple-65B: Small Teams Can Also Build Solid Models
Nathan Lambert commented on Ricursive Intelligence's release of Ripple-65B Preview, calling it an ambitious project and expressing hope that more U.S. labs will follow suit by building solid models with small, focused teams. He noted that the model's approach continues the olmo line, and that seeing a project he was involved in scaled up is meaningful. The quoted tweet states that Ripple-65B is a 65B dense large language model pretrained from scratch, surpassing larger models such as Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B on agentic coding tasks among U.S. open-weight models.
Heat trend for this story
- Current
- 22
- Period peak
- 38 Nov 5 05:00
- 24-hour change
- —
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.