Ripple 65B Preview on BenchGecko · open weights Apache 2.0 dense coding model
Ricursive Intelligence released Ripple 65B Preview on 2026-11-04 under Apache 2.0. BenchGecko lists the 65B dense model at $0.20/$0.60 per M tokens with a 1M token context.
What launched
Ricursive Intelligence released Ripple 65B Preview on November 4, 2026 under Apache 2.0. It is a dense text model trained from scratch on 12 trillion tokens, with a focus on agentic coding. BenchGecko adds the model and a new Ricursive provider today.
Weights are on Hugging Face as ricursiveAI/Ripple-65B-Preview, with BF16 instruct, NVFP4, base and DFlash2 draft variants. Hosted endpoints are live on Hugging Face, Together AI, GMI Cloud and OpenRouter.
Specs
From the Hugging Face model card and Ricursive launch page:
- Parameters · about 65B, dense (all active per token)
- Architecture · thin and deep transformer, d_model 5,120, 80 layers
- Context · up to 1,048,576 tokens
- Inputs and output · text
- Licence · Apache 2.0
- Training · 12T pretrain tokens, mostly English and code
- Serving · BF16 and experimental NVFP4 checkpoints
Pricing on BenchGecko
OpenRouter lists ricursiveai/ripple-65b-preview through GMI Cloud at $0.20 input, $0.60 output and $0.05 cached input per million tokens, with a 1,048,576 token context and up to 131,072 completion tokens. BenchGecko stores those rates on the model row. A task with 1 million input tokens and 500,000 output tokens costs about $0.50 before any cache discount.
Vendor reported benchmarks
Ricursive reports these results. They are self reported and not yet independently reproduced, so BenchGecko records them as vendor claims.
- TerminalBench 2.1 · 80.52
- Comparison on the same card · Reflection Beam 501B 80.1, Meta Muse Spark Open 816B 76.4, Thinking Machines Inkling 975B 63.8, NVIDIA Nemotron 3 Ultra 550B 56.4
Ricursive also states a full R&D cost of $32 million for the Preview release.
Sources
Source: BenchGecko
Written automatically from BenchGecko's tracked data for this period; every number comes from the changelog, model, mindshare and pricing datasets. Checked against the source data before publishing.