A great resource for the Spectrum versions and images is https://spectrumforeveryone.com/technical/zx-spectrum-models/

Below is a quick access visual of model versions to images. All credits from the link above.

The first ever Spectrum produced.

Identical to the Issue 1 from the outside, but often with a blue keymat.

Substantial changes were introduced with these PCB issues, notably the position of the regulator and heatsink (now positioned at the back of the PCB, next to the expansion slot), the colour and clock adjustment controls (TC1/TC2/VR1/VR2) are removed and replaced with an automatic circuit, the beeper is now transistor driven and is a 40ohm impedance part (instead of 200ohm on earlier models) and the ULA type is now 6C001E-6.

These modifications improved reliability somewhat, but unfortunately the 6C001E-6 ULA has a reputation of failing more often than other types.

This issue incorporates the 6C001E-7 ULA type, and also includes earlier service modifications to improve the DC-DC converter reliability directly on the PCB. Some timing issues were presumably discovered during layout, resulting in the /RAS line being routed through spare inverters to delay the signal (this is why Issue 4A and 4B machines won’t work with IC24 removed, unlike earlier models).

The first version to be widely available in the new Spectrum+ casing. Internally the PCBs are identical and you can fit any variant of PCB in either the rubber key or Spectrum+ case.

On the 4B, note the SN94459N at IC14 instead of the usual LM1889N. This was known to cause colour issues.

Almost identical to the Issue 3B PCB (as it doesn’t route /RAS through IC24), but manufactured by Samsung and has green solder mask on both sides of the PCB, and extra solder pads so that alternative components can be fitted (notably the coil, TR4 and TR5). This can be found in both rubber key and Spectrum+ cases with serials starting with S01.

Detailed in the service manuals, only 1000 of these machines were thought to have been produced and none were known to still exist until recently. They have a layout error which requires a PCB modification and patch wiring to work, so presumably this is the reason for such low numbers.

They are also notable for being the first issue to incorporate the ZX8401/PCF1306P IC which was designed to replace the six 74-series logic IC’s on earlier issues. These in themselves caused further timing issues which required the inclusion of an extra 74LS04 on the PCB to fix.

The final 48K PCB issue, this seems to be a bugfixed version of the Issue 5, but also supports the AMI SAGA-1A ULA, which was intended to be an alternative to the Ferranti ULA but DRAM timing issues along with process variation problems led to its cancellation.

Only one Issue 6A machine with an AMI ULA is known to survive.

The Issue 6A also contains dedicated provision on the PCB for a reset switch to be connected, instead of the switch being installed across C27 as on earlier issues in Spectrum+ cases.