Reflection AI has released a preview of its Ripple-65B model, a 65 billion parameter dense large language model trained from scratch. The company, based in Palo Alto, California, reportedly invested $32 million in the training process. The announcement also highlighted other models from different labs, including NVIDIA's Nemotron 3 Ultra-550B and Meta's Muse Spark Open-816B, though the benchmarks presented are noted as potentially cherry-picked. AI
IMPACT This release adds another large-scale model to the growing landscape of LLMs, potentially impacting research and development in the field.
RANK_REASON Frontier-lab model release with system card [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →