It uses similar approach to Norton, that's why they had a product called "Panda Cloud Antivirus". In fact, Panda was the first one to start using "Collective Intelligence" back in 2006.
It doesn't require MOTW, but it doesn't have definitions either, just some hashes of prevalent malware (locally). Upon scanning it will calculate hash, send it to their server and get a result whether the file is safe or not. The server probably doesn't deem brand new files safe and this is where their detection rate comes from. In addition, they partner with Cyren (you can find info on that if you Google them). They get threat intelligence from them and that's why they still manage to achieve good results.
I never managed to see their behavioural blocker in action though. More can be seen here:
Cyber-Security and Technology - Panda Security
Their core technologies are not the worst, but it needs more users/telemetry and the whole software needs a serious redesign.