# Ricursive releases preview of 65B open-weight Ripple model

> Ripple-65B Preview is licensed under Apache 2.0 and scores 80.5 on Terminal-Bench 2.1. Details about access and test conditions were not provided.

Oossa · 2026-11-04 · https://oossa.com/en/ricursive-releases-preview-of-65b-open-weight-ripple-model

Ricursive Intelligence has released Ripple-65B Preview, a 65-billion-parameter language model whose weights are offered under the Apache 2.0 license. MarkTechPost reports that the model scored 80.5 on Terminal-Bench 2.1, a benchmark named in the release coverage. The score is presented as higher than results from US open models as much as 15 times larger.

A model’s weights are the learned values that shape its responses. “Open-weight” means those values are made available, while Apache 2.0 is a license that generally permits use, modification and redistribution subject to its terms. The announcement identifies the license, but does not say where people can get the model, what hardware it requires, or whether there are restrictions beyond the license.

## What does the score show?

The 80.5 figure is the clearest performance detail provided. The comparison with models up to 15 times larger is potentially useful for readers interested in how much computing capacity a model may need, but the report does not name those models or give their scores. It also does not specify the benchmark settings, who ran the evaluation, or whether the results have been independently checked. That leaves the comparison difficult to assess on its own.

The “dense” label describes a model design in which the full set of model parameters is used for each step of generating an answer, rather than selecting only part of the model. The announcement does not provide further technical details or explain how Ripple-65B performs on tasks beyond Terminal-Bench 2.1. It is also called a preview, so the release should not be treated as a statement about a final version or a promise of performance in everyday use.

## What is still unclear?

Ricursive’s release adds another open-weight model to a field where developers can inspect and run model weights under stated license terms. But the details needed to judge practical use are limited here: there is no stated price, download location, hardware guidance, or independent test information. The reported score offers one benchmark result, not a general measure of how well the model will write, answer questions or handle software work.

## The facts

- Ricursive Intelligence released Ripple-65B Preview, described as a 65-billion-parameter dense model.
- The model’s weights are licensed under Apache 2.0.
- MarkTechPost reports a score of 80.5 on Terminal-Bench 2.1.
- The report says the score is ahead of US open models up to 15 times larger.

## Why it matters

Developers who want to experiment with model weights may be able to use, modify or redistribute Ripple-65B under Apache 2.0’s terms. But the available details do not establish how to access it, what it costs to run, or whether the benchmark comparison holds under independently checked conditions.

## Sources & references

1. [Ricursive AI Releases Ripple-65B Preview: A 65B Dense Open-Weight Model Scoring 80.5 on Terminal-Bench 2.1](https://www.marktechpost.com/2026/11/04/ricursive-ripple-65b-preview-open-weight-model/) – MarkTechPost, 2026-11-04

Last updated: 2026-11-04