At the recent Q2B conference, IonQ announced that they will initiate a quantum cloud computing service to select beta users in 2019. Additionally, they described some of the prototype test results on some of their early hardware.
IonQ provided some details on their gate fidelities on an 11 qubit implementation that shows average single qubit fidelities greater than 99% and double qubit fidelities greater than 98%. These numbers are better numbers than currently achieved by the superconducting qubit implementations. They also showed a benchmark of the Bernstein-Vazirani algorithm that showed a 73% success rate for a 10 qubit oracle versus an estimated 0.2% success rate if this algorithm was performed on a classical computer.
IonQ still has some additional work to do in scaling up their technology. They showed some preliminary indications that 79 qubit and even 160 qubit implementations are possible, but apparently they have not fully tested these configurations yet.
One advantage of the IonQ ion trap technology versus superconducting technology is that it provides all-to-all connectivity. This means that one can form a two-qubit gate between any two pairs of the 11 qubits. This makes it easier to map a program onto the hardware as the connectivity of the superconducting technologies is more limited. On the other hand, ion trap technology has much slower gate delays than the superconducting versions. Typical gate delays are on the order of about 50 microseconds for the ion traps versus about 50 nanoseconds for the superconducting technology.