Basically one lane (single-lamba) versus 4 as in LR4 and it's Cisco's approach to simplication vs POET's approach with the IO.

Lots to read on the subject:

https://blogs.cisco.com/networking/whats-so-great-about-lr-compared-to-lr4-optics

https://blogs.cisco.com/sp/how-to-make-100g-pluggable-optics-in-massive-volume

Its an in-house blog rom Pat Chou Product Manager - Transceiver Modules Group

vs.

A 'Market Perspective' from Vivek.

* https://poet-technologies.com/news/2021-apr-22.html

Remember, POET builds 'optical engines'! The 100G LR/LR4 are "products" that use embedded components - like the OI.

The real question is:

Who is embedding the POET OI to produce the 100G LR4 modules?

IMHO it's an apples and oranges discussion.

Also:

https://www.arista.com/assets/data/pdf/Arista100G_TC_QA.pdf shows both LR4, LR4L and LR offerings.