Ricursive Intelligence Open-Sources Ripple-65B Preview Coding Model Under Apache-2.0
The dense 65B agentic coding model ships under Apache-2.0 with BF16 and NVFP4 checkpoints; benchmark figures remain vendor claims.
Ricursive Intelligence released Ripple-65B Preview on 4 November 2026, an open-weight language model for agentic coding published under the Apache-2.0 license without additional restrictions. The model is a dense transformer with 65 billion parameters across 80 layers, trained on 12 trillion tokens, and supports a context length of up to 1 million tokens. Checkpoints are available in BF16 and NVFP4 formats, alongside a base model, and community GGUF quantizations appeared within two days of release. The weights repository is live on Hugging Face.
According to the vendor's model card, the model scores 80.52 on TerminalBench 2.1 and 79.1 percent on SWE-Bench Pro v2.0 Hard; these benchmark figures are vendor claims and have not been independently verified. Independent coverage noted the model trails Inkling 2 Small on several coding and agent benchmarks.
The model is served through provider platforms at a listed price of USD 0.20 per million input tokens and USD 0.60 per million output tokens.