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

I am running qwen3.8 27b on my dual rtx 3090 (fp8 quant, unquantized cache, 129k context) and I think it works decently well with hermes, opencode etc. But! I am tempted by the new models coming out such as qwen3.8 flash next, deepseek v4 flash, glm 5.3 flash. However, there is a big jump in vram and therefore in cost! The least expensive option seems to be buying 2 of those cmp 170hx 64gb cards for roughly 6-7000 usd in total (that is the price I can find for verified cards here in europe at least). With that I would get another 128gb of vram for a total of 176gb so I could run I think around 3-4bit quants of the above models, right?). I am thinking that it might be faster because of moe but not sure how much smarter? For that kind of money I would want a real noticable improvement! 4xv100 32gb would be cheaper (maybe half price?), but even more hassle to set up, more power draw, and slower. What do you think? The free option is to just wait for qwen4 27b and (hopefully) just download more IQ.