Ricursive Says Ripple-65B Preview Scored 80.52 on Coding Benchmark
The 65-billion-parameter open-weight model supports context windows of about 1 million tokens. Ricursive put research and development costs at about $32 million.

Ricursive Intelligence said its Ripple-65B Preview scored 80.52 on TerminalBench 2.1, exceeding the score of a larger model on the agentic coding benchmark.
Ricursive compared the result with Reflection Beam 501B, a U.S. open-weight model with more parameters that scored 80.1. The company said Ripple-65B Preview is a dense, 65-billion-parameter model pretrained from scratch.
The model supports context windows of about 1 million tokens. Ricursive put its total research and development costs at about $32 million.
The model’s weights are available through Hugging Face, Together AI, GMI Cloud and OpenRouter.