[Ricursive Releases Open-Source Weight Model Ripple-65B Preview] According to a TechFlow report on November 5th, Ricursive Intelligence officially announced the release of its open-source weight model, Ripple-65B Preview. This model employs a dense architecture with 65 billion parameters, was pre-trained from scratch, and supports contexts with up to 1 million tokens. Ricursive claims that the model achieved a score of 80.52 on the TerminalBench 2.1 benchmark for agent programming, surpassing Reflection Beam 501B's 80.1. The company stated that the total R&D cost for the model's development and training was $32 million, and the weights are already available on platforms such as Hugging Face.