MarkTechPost:AMD Releases Instella-MoE-16B-A3B: A Fully Open Mixture-of-Experts LLM With 2.8B Active Para…
原文摘要:AMD released Instella-MoE-16B-A3B, a fully open Mixture-of-Experts language model trained from scratch on Instinct MI300X and MI325X GPUs. It holds 16B total parameters but activat 来源:MarkTechPost。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。