As of today, bcrypt, scrypt and PBKDF2 offer reasonable properties to be used and considered safe.
What is especially on their side is their history. As Neil pointed out in the comments, Argon2 - while winning the PHC - still has to prove itself, being researched extensively to be fairly sure nothing has been overlooked that can break its neck.
That's the reason established hash functions (and encryption systems) are preferred over new ones as long as they are still not feasible to break: They have been thoroughly researched and proven themselves over many years.
The people over at crypto.SE have some questions on Argon2 that may answer some of the more it-depth questions regarding comparisons and Argon2 itself.
Usually, it takes a few years for a new algorithm to become generally accepted after release - that's because reliable reviews require a great amount of research and time and the academic processes often involved are lengthy.