In this paper, we investigate adaptive transmission for multiple-input multiple-output (MIMO) antenna systems with multi-beams in fading environments when the channel-state information at both the transmitter and the receiver is available. In Rayleigh fading environments, the transmitter is supposed to be capable of spatiotemporal subchannel selection and power control as eigenvalues of channel matrix changes. Under constraints of individual bit error rate (BER) and maximum transmit power for each data stream, we adopt the optimal transmit strategy of minimizing the average transmit power (ATP), and focus on the case where all of these individual BER constraints are of short term. With the help of an order statistical result of the eigenvalues of complex central Wishart matrices, we derive and give closed-form ATP expressions. Due to short term BER constraints, adaptive transmission based on channel eigenvalues can keep BER reliability of MIMO system unchanged almost without outage, just working in additive white Gaussian noise environments. When the maximum transmit power is allowed to be infinite, we analyze deeply relationship among the ATP, the numbers of transmit and receive antennas. Finally, some numerical results are provided to validate the theoretical analysis and make comparisons with the corresponding adaptive transmit scheme under long term BER constraints. Our simulation results show that the adaptive transmit scheme of short term is attractive for the future MIMO applications, especially when a very large MIMO system is employed.