A Reddit user reports successfully running Qwen3.8-Flash-Next FP8 locally at approximately 60 tokens per second using an RTX 5090 with 128 GB DDR5 and the Infernix engine, streaming experts from NVMe. The author expresses strong anticipation for Qwen 4.0 and suggests that investing in high-end hardware is justified for daily users currently reliant on cloud AI services.