If you guys want any kind of Snapdragon, the device will likely get significantly more expensive…
Anyway, you have 2x2 options:
First, you can choose between their Mobile line (the kind of chips that go into smartphones, like 888 or 865) and their Compute line (the ones that go into WoS laptops and chromebooks). Personally, I’d vouch for the former, as the Compute chips are usually a gen or two behind compared to their Mobile counterparts and the performance gains don’t really make up for their higher TDP.
Then you can choose between the 8xxx and 7xxx line. The former is more powerful (though mainly on the GPU and coprocessors [think accelerators, DSPs and camera stuff] front nowadays), but also has noticeably higher price and heat output.
7xxx line SoCs are perfectly fine for a snappy experience, they often have a well-performing CPU configuration (though, of course not as powerful as their higher-up counterparts) with a midrange GPU option. They seem to never really overheat though, at least in normal use. While I didn’t yet get to play with the newest SM7325/7350 devices that were announced this year, the specs on paper make it look like something to consider… Of course Jingling probably doesn’t have the kind of money to manufacture the devices in an instant, so we’d have to wait a generation or two (nobody likes buying tech that’s outdated on release + Qualcomm likes to EOL their products seemingly randomly). But if the current gen looks suitable already, the next ones should only be better.
Then, comes the software problem. If you want to go with a Qualcomm downstream kernel, you’ll (likely) have to put up with many compromises, especially with the amount of binary packages you have to ship to even have the thing working properly. If you go with the mainline kernel though (oh and did I mention I do qcom mainline stuff? shameless plug ;)) you’d either have to follow the PinePhone route, where the community or select developers take the task of getting as many things working as possible (having full 100% absolutely everything works all the time functionality is proooobably out of reach for the time being), or fight with mainline code yourself and take on the responsibility for both making sure it works as you ship the product AND upstreaming patches to the mainline kernel, so that you don’t essentially create your own downstream. Adding to that, developing on Qualcomm platforms is not always pretty - if Jingling wanted people to help porting the mainline kernel, they’d have to provide a way of emergency flashing the device (via a vendor-signed EDL loader), or not blows security fuses on their devices, making them very developer-friendly, but not safe at all from attacks for the regular customer.
If you have any more questions about what going qcom would imply, I can try to answer, as I guess I know a thing or two…