November 20, 2026

Ricursive AI Debuts Ripple-65B Preview, Scores 80.5 on Terminal-Bench

A new 65-billion-parameter dense open-weight model from Ricursive AI posts a notable Terminal-Bench 2.1 score, according to Marktechpost.
Ricursive AI Debuts Ripple-65B Preview, Scores 80.5 on Terminal-Bench

Ricursive AI has released a preview version of Ripple-65B, a 65-billion-parameter dense open-weight language model, Marktechpost reported.

According to the outlet, the model achieved a score of 80.5 on Terminal-Bench 2.1, a benchmark used to evaluate model performance on terminal and command-line based tasks. The release is being positioned as an open-weight alternative in a field increasingly dominated by proprietary systems from major AI labs.

Marktechpost’s report did not provide extensive detail on the model’s training data, architecture specifics beyond its dense 65B parameter count, or licensing terms. The designation as a preview suggests Ricursive AI may continue to refine or expand the model before a full release.

Dense models of this scale, as opposed to mixture-of-experts architectures, typically require substantial compute resources to run but can offer more predictable performance characteristics across tasks. The Terminal-Bench 2.1 score places Ripple-65B among models capable of handling complex, multi-step command-line and agentic workflows, a category of benchmark that has gained attention as developers test models on practical coding and systems administration tasks rather than purely conversational ability.

The release adds to a growing list of open-weight models from smaller and emerging labs seeking to compete with closed offerings from larger organizations. Open-weight releases allow researchers and developers to inspect, fine-tune, and deploy models without relying on API access from the original creator, a factor that has fueled interest from the research community and enterprises seeking more control over deployment.

Further technical details, including training methodology and broader benchmark comparisons, were not specified in the original report. Marktechpost’s coverage did not indicate a timeline for a non-preview release of Ripple-65B.

Based on reporting by www.marktechpost.com.

Leave a Reply

Your email address will not be published. Required fields are marked *