Ricursive releases open-weight Ripple-65B, claiming US open-weight SOTA on TerminalBench 2.1
StanfordAILab · x · 2026-11-12
Ricursive Intelligence released Ripple-65B Preview, an open-weight model pretrained on 12 trillion tokens with up to 1M context length. Its dense architecture targets low-latency inference and local deployment, running efficiently on devices like the NVIDIA DGX Spark.
On agentic coding it hits 80.52 on TerminalBench 2.1, a US open-weight SOTA, surpassing much larger models: Reflection Beam 501B (80.1), Meta Muse Spark Open 816B (76.4), Thinking Machines Inkling 975B (63.8), and NVIDIA Nemotron 3 Ultra 550B (56.4). The company claims roughly one-tenth the training cost of comparable frontier models. Available on Hugging Face, Together AI, OpenRouter, Fireworks and GMI Cloud.
More from Models
- "Anthropic spent $10B+ on Opus 5.5" is fake news, per community debunk — ns123abc · 2026-11-12
- Domingos: frontier AI capability is rising linearly, +16.5 ECI points/year, no exponential — pmddomingos · 2026-11-12
- Hark launches Pro⁴ plan for power users with 10x the usage of Pro² — adcock_brett · 2026-11-12
- Why Is Inkling 2 Small Trained on So Few Tokens Yet Still Competitive? — georgejrjrjr · 2026-11-12
- Dev benchmarks: OpenCode trails Pi as DeepSeek dominates agentic coding quadrant — mariofilhoml · 2026-11-12
- RuneBench cost-vs-score chart shows DeepSeek dominating the budget quadrant — maxbittker · 2026-11-12