- Support for NVIDIA K40 and K80 GPUs on Linux systems, for HF and DFT calculations.
- Speedups for key parts of several calculation types.
- CASSCF improvements and support for active spaces of up to 16 orbitals.
- An optimized memory algorithm which avoid I/O during CCSD iterations.
- Enhancements to GEDIIS optimization algorithm.
- CASSCF improvements for calculations using active spaces larger than (10,10). Active spaces of (16,16) can noe be feasible (depending on the system).
- Faster performance for the low level part of the OVGF compound model for predicting photoelectron spectra.

Gaussian 16 also offers a wide variety of performance improvements to many different parts of the program. Among the most significant are the following:

For example, a **DFT Opt=CalcFC Freq** calculation on Alanine 254 is 1.8 times faster in Gaussian 16 than in Gaussian 09. The two second derivative calculations for this job comprise 47% of the execution time in Gaussian 16 compared to 61% in Gaussian 09.