Ripple-65B Preview

Ripple-65B Preview is an open-weight language model from Ricursive Intelligence, a lab that builds AI for chip design. The company says it is a general-purpose variant of the proprietary model behind its chip design platform, trained from scratch on 12 trillion tokens, and released on 4 November 2026 with support for up to one million tokens of context.

## A dense model aimed at agentic coding

Unlike most recent large open-weight releases, Ripple is dense rather than a mixture of experts: all 65 billion parameters are used for every token. Ricursive argues this gives strong reasoning with low-latency generation, and says the small parameter envelope lets it run on local machines such as NVIDIA DGX Spark. It reports a score of 80.52 on TerminalBench 2.1 and compares it with much larger open-weight models; that figure is the company's own and is not stored as a score here.

Ricursive also states that the full research and development cost of the model, from the first pretraining experiments to the final run, was $32 million.

## Where to get it

The weights are on Hugging Face under the ricursiveAI organisation, tagged Apache 2.0, with an NVFP4 quantised copy alongside. Ricursive lists Together AI, GMI Cloud, OpenRouter and Fireworks as hosted routes. OpenRouter shows a one-million-token context at $0.20 per million input tokens and $0.60 per million output tokens, with native tool calling.

## Key facts

**Developer:** Ricursive Intelligence **Released:** 4 November 2026 **Size:** 65 billion parameters, dense **Training data:** 12 trillion tokens (company figure) **Context:** up to 1 million tokens **Licence:** Apache 2.0, per the Hugging Face repository **Hosted price on OpenRouter:** $0.20 per million input tokens, $0.60 per million output tokens

Not yet independently verified: Not yet independently verified. The TerminalBench 2.1 result and the training cost are Ricursive Intelligence's own figures; we found no independent test, and the licence is taken from the Hugging Face repository tag. We will update this when it can be confirmed, and remove this note.