I would say a 158 as it is over a 3 hour journey end to end. The 166s were designed as London commuter trains and they aren't great on long distance regional operations while this is the sort of thing the 158s were designed for. They've also got quite a long dwell at Bristol anyway.
GWR operate some services that are over 4 hour end to end (to Weymouth, Great Malvern etc). However all their Wessex services tend to operate as mix of local services, and fast regional services, where some stations are quite widely spaced. They also tend to have high churn levels at places like Bath Spa, and other stations.
None of the 158, 166, 166 fleet is ideal for this mix. The end doors on 158s struggle with the churn, the 165s were never designed for journeys over about an hour, the 166s often end up coupled to a 2car 165 on busier trains so don’t really fit in either. Because many of the local stations are not staffed at times, coupling non-gangwayed stock where guards and ticket inspectors can’t access all passengers, or short platforms isn’t helpful either.
They really need a 5car (or 6-8 section articulated unit), about 120m long bi-mode with dual voltage capability, good number of doors, decent seating (and I mean comfortable for 2+ hour journeys), ideally with level boarding for the longer regional trains. I would say 110mph on 25kv, but around 95mph on 750v and diesel should be ok. For simplicity will call this a regional bi-mode.
The 165s and 166s could be retained for some local trains until early 2030s, (although places like Cornwall would need some staff training), but whatever is introduced would also need training.
Actually GWR are not the only operator that needs the regional bi-mode. There is a strong case for some of its services through Cotswolds too, and even those where IETs are taking over from Castle HSTs (which don’t need a train designed for high speed).
There is a huge gap between high speed intercity trains and local service diesel and commuter EMUs where nothing since the 30+ year old 158, 159 fleet has tried to fill the requirement. If a leasing company speculatively built 50 of these, they would easily find homes in mid 2020s as the sprinter era trains need cascading out. However unlike when Angel ordered the 25 Siemens EMUs speculatively, the DfT now micro manages so speculative orders are now high risk even when a demand is clearly looming.