Ricursive Intelligence says Ripple-65B Preview leads open U.S. models on TerminalBench 2.1 at $32 million total cost
🤖 AI-generated content — The title and summary were produced automatically by artificial intelligence, without human editorial review.
AI chip firm Ricursive Intelligence pretrained Ripple-65B Preview, a 65-billion-parameter dense Nvidia-oriented Transformer, on 12 trillion tokens. It claims an 80.52 TerminalBench 2.1 coding score—the strongest among open U.S. models—with Terminus 2 and OpenCode harnesses. Other benchmarks trail the newly released Inkling 2 Small, though Ripple may be simpler to run on premises. Ricursive states development and training cost $32 million; weights are on Hugging Face under Apache 2.0.
Continue in the app — vote & join in ➔Source: The Decoder (AI) · via ahirlevel.hu