These convective differences between models at initialization/very short-term quickly grow upscale (with time). These discrepancies likely have a significant impact on how strong of a system EPS and GEFS are showing down the road for 95L, in addition to the differences in the operational runs.
Note that the models have switched places since yesterday (when tweet from Eric Webb below was written), with GFS now more convectively active while the ECMWF is more anemic.
GFS has tended to handle relatively small systems developing in the monsoon trough or ITCZ better than the ECMWF in the past few years- for example, Beryl 2018, Dorian 2019, Gonzalo 2020 and Elsa this year in their genesis phase.
https://www.twitter.com/yconsor/status/1426169852050526208
https://www.twitter.com/webberweather/status/1425706495191126021