Ripple-65B dense model from scratch beats 975B giants on agentic coding
afurgs · x · 2026-11-05
RicursiveAI released Ripple-65B Preview, a dense LLM pretrained from scratch that claims US open-weight SOTA on agentic coding, beating much larger models like Inkling-975B and Nemotron 3 Ultra-550B. Commentary explains the intuition: dense params serve reasoning rather than memory — Ripple wins on Terminal-Bench (80.5 vs Beam's 80.1 at 2.8x compute per token) but loses on knowledge tasks (SimpleQA 10.1, MMLU 83.7). It fits on a single RTX Pro 6000 Blackwell at NVFP4.
More from coding & agent
- GitLab CPMO: code is abundant, trust is the bottleneck for AI coding — TechNadu · 2026-11-05
- Dev replaces Conductor with Amp: support, SEO, iOS releases all agent-handled — iannuttall · 2026-11-05
- Open-source PageIndex ditches vectors for tree-search RAG, scores 98.7% on FinanceBench vs GPT-4o's 31% — aakashgupta · 2026-11-05
- Most agent failures are harness failures, not model failures, argues Reddit thread — North_Sugar_15 · 2026-11-05
- Revolut self-hosts GLM on 56 B200s, cutting coding token cost to ~$0.30/M — alex_verem · 2026-11-05
- "Hand-writing code will sound like a fever dream to our children," dev predicts — tlakomy · 2026-11-05