New KNITRO 9.0 features:

• New tuner that automatically detects the optimal parameter settings for a given problem
• New MATLAB interface supporting mixed-integer problems, complementarities, parallelism
• Performance improvements in the Python interface and added NumPy support
• Updated Intel MKL BLAS libraries
• Added support for the Intel MKL PARDISO linear solver (incl. parallel option)
• Improvements to automatic finite-difference computations
• Improvements to stopping criteria for better precision
• New API function allowing users to define individual feasibility tolerances, for each constraint and variable
• New default initial point generation strategy
• Significant speedups on nonlinear programs
• Significant speedups on mixed-integer nonlinear programs
• Significant speedups on mathematical programs with equilibrium constraints (MPCC/MPEC)

According to recent benchmarks, KNITRO and Xpress-NonLinear (includes KNITRO) deliver the best performance among all nonlinear solvers. The results obtained demonstrate KNITRO's superiority, both in terms of speed and robustness.