USB wifi adapters are problematic by design. USB 2.0 devices will just be slow, but USB 3.0 actively emits noise in the 2.4 GHz band. Even the USB Implementers Forum acknowledges this:
Especially with TP-Link devices, the hardware revision is just as important as the model.
I've been using an Archer C7 hardware revision 4.0 for the last few years, and it works fine with OpenWRT.
However, I'd recommend buying a device where the serial console works without having to solder pull-up resistors to the PCB to allow for recovery should you ever come across a bad OpenWRT update.
https://forum.openwrt.org/t/solved-serial-console-tp-link-archer-c7-v4-not-working/112154

I usually don't have any problems connecting to the UART port of my devices but my Archer C7 v4 is giving me problems. From looking at the Wiki / ToH pages version 5 of this router doesn't have the RX connected. It seems that version 4 has something similar. I could find R27 (which is also referenced on version 5), but I still can't get serial data out of the device. Anyone any suggestions on how to get serial console to work on Version 4?
Something like this might give you more bang for the buck (and ships with OpenWRT out of the box):
Note that even though the OpenWRT doesn't mention the device having a serial/UART port, it still seems to have one:
https://github.com/openwrt/openwrt/commit/3307fe8ee4cf8ee52cefd03feb97f1d826253eeb
Also, with WiFi 6, 128 MiB of flash and 512 MiB of RAM, this is a lot more future proof than the Archer C7 (wihich only features WiFi 5, 16 MiB of flash and 128 MiB of RAM).
Wifi antennas are not genitalia, size doesn't necessarily correlate with functionality.
I don't claim to be an RF expert either, but my day job involves a huge eduroam deployment with several thousand access points, and none of them have protruding wifi antennas, and I do have a ham radio license.
Yes, in theory, the 2.4 GHz radio waves signals a wave length of about 12.5 cm, and a 5 GHz radio signals have a wave length of about 6 cm.
In practice however, you won't be able to align
four external antenna in an optimal way anyway, and even if you would, it wouldn't matter because WiFi is used with mobile devices that are carried around, held in sweaty hands, their signals attenuated by walls, reflected by mirrors and windows with a heat-reflective metal coating on their glass panes, etc.
Modern WiFi devices use multiple antennas/physical beams, and both the AP and the client device dynamically figure out the parameters for an optimal result.
And this "figuring it out dynamically" works best if at least the relative positions of the device's antennas to each other are fixed. Also, internal antennas will usually be either integrated into the PCB traces or soldered directly onto the PCB and can be calibrated properly when designing or manufacturing the device.
I'd expect external detachable antennas to incur additional attenuation through the SMA connector and the shitty internal coax pigtail cable.
Also, aside from all these theoretical musings, the data sheets are pretty clear in this case.
GL-AX1800:
> Wi-Fi Speed: 600Mbps (2.4GHz), 1200Mbps(5GHz)
GL-B3000:
> Wi-Fi Speed: 574Mbps (2.4GHz), 2402Mbps (5GHz)
So I'd definitely choose the GL-B3000 over the GL-AX1800 anytime if you want good wifi performance.
The AX1800 does have one advantage though: It has an SoC with 4 CPU cores instead of just 2, and those 4 cores are clocked higher, allowing for more VPN throughput.
Also, the GL-B3000 uses only 13 W of power, the GL-AX1800 uses 18W.
If you are wondering why the GL-AX1800 is slower with 5 GHz: The 5 GHz radio in that device is a QCN5052, which can only handle 80 MHz channels, and each 5 GHz antenna/beam gives you up to 600 MBit/s with 1024QAM.
The GL-B3000 uses a QCN6102, which can handle 160 MHz channels, thus doubling theoretical throughput per antenna/beam.
Both the 2.4 GHz and 5 GHz radios in both devices can only handle 2 antennas btw.
For further details, about the SoCs and radios/phys used, I can recommend this website:
https://wikidevi.wi-cat.ru/Qualcomm
Btw, depending on the area/rooms to be covered and given the fact that 5 GHz is much more susceptible to being attenuated by walls and stuff, a budget of $ 100 is likely better spent on two faster 5 GHz APs than on a single AP with an inferior 5 GHz radio and an only marginally better 2.4 GHz radio.
There's legitimate use cases for both devices.
For example, the GL-AX1800 might be a better choice for somebody who is operating a "Freifunk" WiFi in Germany, where all traffic is usually tunneled through Wireguard or some other VPN to a dedicated legal entity (a local non-profit association that cosplays an ISP to shield AP operators from legal liabilities because most legalese people can't be bothered to understand how free WiFi works).
As far as radio throughput goes, expect somewhere between half and a third of the data rates claimed in the data sheet. The datasheet numbers are what's possible under ideal conditions (basically "if client and AP were not actually using antennas but their antenna ports were connected directly to each other using high-quality coaxial cable, in an environment with no EMI").
Also, don't be fooled by excessively high throughput figures in the datasheets of APs that have more then two antennas per frequency band/radio. Real-world client devices will only have one or two antennas anyhow, therefore such APs are only useful in scenarios with lots of client devices, for example big meeting or class rooms.
@dalias @srslypascal yes: the Flint 2 runs a notably more recent version of OpenWRT (21.x) than the AX-1800/Flint 1², plus it runs mainline OpenWRT (24.x)³. (Avoid the Flint 3: it runs Qualcomm hardware and works poorly with mainline OpenWRT⁴.)
(That said, the Flint 2 does cost a good chunk more than the Flint 1. That might be a disqualifier.)
¹: https://dl.gl-inet.com/router/mt6000/stable
²: https://dl.gl-inet.com/router/ax1800/stable
³: https://dl.gl-inet.com/router/mt6000/openwrt24
⁴: https://www.servethehome.com/gl-inet-flint-3-be9300-wifi-7-2-5gbe-router-review-good-in-many-ways/
⁵: https://www.servethehome.com/gl-inet-gl-mt6000-flint-2-wifi-router-review-mediatek-openwrt/
The whole point of getting a device that is supported by mainline OpenWRT is being able to use it with mainline OpenWRT without relying on the hardware vendor for firmware updates.
https://firmware-selector.openwrt.org/?target=qualcommax%2Fipq60xx&id=glinet_gl-ax1800
https://firmware-selector.openwrt.org/?target=qualcommax%2Fipq50xx&id=glinet_gl-b3000
TBH I don't use the web interface myself, and I compile my own custom images for my devices.
AFAIK, OpenWRT has switched from opkg to apk as the default package manager with OpenWRT 25.x, and the default images should contain the "luci" package, which is the web interface.
What exactly do you mean by "tiny"? AFAICT, both the "sysupgrade" and "factory" versions of the firmware image should be somewhere around 14 MiB in size and have (among other things) the luci package pre-installed.