RELEASE · MODELS · #3242
Ricursive releases Ripple-65B Preview — 65B Apache 2.0 US model with strong coding score
Ricursive Intelligence released Ripple-65B Preview, a dense 65-billion-parameter Transformer pretrained (company says) on 12 trillion tokens and reported to score 80.52 on the TerminalBench 2.1 coding benchmark — the company describes this as the best result among open US models. The model, trained (company says) for about $32 million, uses Terminus 2 and OpenCode as its harness, is optimized for Nvidia hardware, and is available on Hugging Face under an Apache 2.0 license; Ricursive notes Ripple trails the newly released Inkling 2 Small on some benchmarks but may be easier to run on users' own hardware.
KEY POINTS
- Ricursive Intelligence released Ripple-65B Preview, a dense 65-billion-parameter Transformer pretrained (company says) on 12 trillion tokens and reported to score 80.52 on the TerminalBench 2.1 coding benchmark — the company describes this as the best result among open US models.
- The model, trained (company says) for about $32 million, uses Terminus 2 and OpenCode as its harness, is optimized for Nvidia hardware, and is available on Hugging Face under an Apache 2.0 license; Ricursive notes Ripple trails the newly released Inkling 2 Small on some benchmarks but may be easier to run on users' own hardware.
- An Apache 2.0 65B model with strong reported coding performance matters because it expands permissively licensed, on-premise-capable alternatives in the open-US model landscape.
WHY IT MATTERS
An Apache 2.0 65B model with strong reported coding performance matters because it expands permissively licensed, on-premise-capable alternatives in the open-US model landscape.