Developer loaded 670GB of Kimi K2.6 across a 4-node Apple rack. cut his AI bill from $4,000 to $700/month 2 nodes: 23.4 tokens/sec. 4 nodes: 28.9 tokens/sec. 100% GPU on every machine. MLX + RDMA scaling works > Opus 4.8: plans. writes spec. catches flaws. $5/M tokens > Kimi
Developer loaded 670GB of Kimi K2.6
Cut AI costs from $4,000 to $700/month while increasing processing speed.