Worth going from Qwen3.8 27B to flash next or maaybe deepseek v4 flash?

A user running Qwen3.8 27B on dual RTX 3090s is considering upgrading to newer models like Qwen3.8 Flash Next or Deepseek V4 Flash. The user is weighing the cost of acquiring two CMP 170HX 64GB cards for 6-7000 USD to achieve 176GB VRAM, which would allow for 3-4bit quants of these models. While this might offer speed improvements due to MoE, the user seeks a noticeable intelligence improvement for the investment, also considering 4xV100 32GB as a cheaper but more complex alternative.

This post offers a rare, detailed comparison of the cost and performance considerations for upgrading from Qwen3.8 27B to newer Flash models, unlike typical discussions that focus solely on model capabilities.