Visualizations of Mittelmann benchmarks

Interactive plots showing pairwise time differences for every instance and every solver

Interactive charts comparing the results of Hans Mittelmann’s benchmarks. Each solver can be selected to show pairwise running time factors for every other solver in the respective benchmark. These plots should make browsing the results easier. The score (scaled shifted geometric mean) is recomputed using the reported solving times. We also calculate a “virtual best” or “ensemble” solver that picks the best performance over all solvers for every single problem instance. This might reveal how much potential the individual solvers still have. Please let me know if you have a question or if there is an error.

LPfeas Benchmark (find PD feasible point) (9 Dec 2022)

Choose base solver for comparison:

solver score (as reported) solved of 71
⭐ virtual best 0.64 100%
🥇 Gurobi-10.0.0 1.00 (1.00) 100%
🥈 COPT-6.0.3 1.04 (1.04) 100%
🥉 MOSEK-10.0.25 2.01 (2.01) 100%
📊 ORTOOLS-9.4 15.10 (15.10) 77%
📊 HiGHS-1.4.1 18.23 (18.20) 82%
📊 KNITRO-13.0.0 20.85 (20.80) 70%
📊 MATLAB-R2022b 27.38 (23.70) 79%
📊 Tulip-0.9.4 70.61 (70.60) 56%

LPopt Benchmark (find optimal basic solution) (25 Jan 2023)

Choose base solver for comparison:

solver score (as reported) solved of 71
⭐ virtual best 0.81 100%
🥇 COPT-6.0.3 1.00 (1.00) 100%
🥈 Gurobi-10.0.0 1.18 (1.18) 99%
🥉 Optverse-0.2.13 3.65 (3.65) 90%
📊 MOSEK-10.0.24 5.14 (5.14) 82%
📊 HiGHS-1.4.1 16.89 (16.90) 79%
📊 CLP-1.17.7 28.99 (29.00) 61%
📊 MATLAB-R2022b 41.14 (41.10) 65%
📊 Google-GLOP 60.65 (60.70) 48%
📊 SOPLEX-6.0.0 98.11 (98.10) 48%
previous benchmarks 🔽

Large Network-LP Benchmark (commercial vs free) (24 Jan 2023)

Choose base solver for comparison:

solver score (as reported) solved of 25
⭐ virtual best 0.88 100%
🥇 OptVerse-0.2.13 1.00 (1.00) 100%
🥈 MindOpt-0.18.6 1.20 (1.20) 100%
🥉 Gurobi-10.0.0 1.60 (1.60) 100%
📊 COPT-5.0.0 2.02 (2.02) 100%
📊 Clp-1.17.7 5.66 (5.66) 100%
📊 HiGHS-1.4.1 11.91 (11.90) 80%
📊 MATLAB-R2022b 19.84 (19.80) 80%
📊 MOSEK-10.0.24 24.46 (24.50) 88%
📊 QSopt-1.01 33.08 (33.10) 68%
📊 SOPLEX-6.0.0 62.91 (62.90) 64%
previous benchmarks 🔽

The MIPLIB2017 Benchmark Instances - 8 threads (13 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 240
⭐ virtual best 0.82 96%
🥇 Gurobi-10.0.0 1.00 (1.00) 95%
🥈 COPT-6.0.1 2.63 (2.63) 82%
🥉 SCIPC/spx-8.0.0 8.92 (8.92) 63%
📊 HiGHS-1.3.0 9.28 (9.27) 65%
📊 SCIP/spx-8.0.0 10.90 (10.90) 57%
📊 CBC-2.10.5 16.30 (16.30) 45%
previous benchmarks 🔽

MILP cases that are slightly pathological (26 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 45
⭐ virtual best 0.82 98%
🥇 GUROBI-10.0.0 1.00 (1.00) 98%
🥈 COPT-5.0.0 4.79 (4.79) 76%
🥉 HiGHS-1.2.2 20.74 (20.70) 56%
📊 SCIPC-8.0.0 20.80 (20.80) 51%
📊 SCIP-8.0.0 26.66 (26.70) 42%
📊 CBC-2.10.7 41.62 (41.60) 11%
📊 GLPK-5.0 42.19 (42.20) 13%
📊 MATLAB-2020a 50.84 (50.80) 7%
previous benchmarks 🔽

Infeasibility Detection for MILP Problems (22 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 32
⭐ virtual best 0.92 94%
🥇 GUROBI-10.0.0 1.00 (1.00) 94%
🥈 COPT-6.0.1 1.62 (1.62) 94%
🥉 SCIPC-8.0.0 6.24 (6.24) 81%
📊 SCIP-8.0.0 8.00 (8.00) 78%
📊 HiGHS-1.2.2 10.72 (10.70) 91%
📊 CBC-2.10.5 19.53 (19.50) 62%
📊 MATLAB-2020b 30.36 (30.00) 47%
previous benchmarks 🔽

Several SDP-codes on sparse and other SDP problems (4 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 76
⭐ virtual best 0.59 99%
🥇 COPT-6.0.1 1.00 (1.00) 99%
🥈 MindOpt-0.22.0 1.77 (1.80) 97%
🥉 MOSEK-10.0.27 2.99 (3.07) 95%
📊 SDPT3-4.0 4.47 (4.62) 91%
📊 CSDP-6.2.0 4.52 (4.68) 92%
📊 HDSDP-0.9.2 7.31 (7.62) 92%
📊 SDPA-7.4.2 8.98 (9.39) 80%
📊 SeDuMi-1.3.5 24.46 (25.90) 82%
previous benchmarks 🔽

Large Second Order Cone Benchmark (12 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 18
⭐ virtual best 0.77 100%
🥇 MOSEK-10.0.18 1.00 (1.00) 100%
🥈 Gurobi-10.0.0 1.11 (1.11) 100%
🥉 COPT-6.0.0 1.15 (1.15) 100%
📊 KNITRO-13.0.0 10.06 (10.10) 83%
📊 ECOS-2.0.4 83.02 (83.00) 33%
previous benchmarks 🔽

Mixed-integer SOCP Benchmark (12 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 47
⭐ virtual best 0.94 100%
🥇 Gurobi-10.0.0 1.00 (1.00) 100%
🥈 COPT-6.0.1 2.78 (2.78) 98%
🥉 MOSEK-10.0.18 14.07 (14.10) 72%
📊 SCIP-8.0.0 24.28 (24.30) 66%
previous benchmarks 🔽

Binary Non-Convex QPLIB Benchmark (13 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 96
⭐ virtual best 0.69 100%
🥇 Gurobi-10.0.0 1.00 (1.00) 97%
🥈 OCTERACT-4.5.1 1.76 (1.76) 95%
🥉 Baron-22.9.1 7.59 (7.59) 57%
📊 SCIP-8.0.0 38.48 (38.50) 36%
📊 ANTIGONE-1.1 78.03 (78.00) 17%
📊 COUENNE-0.5 92.85 (92.90) 6%
previous benchmarks 🔽

Discrete Non-Convex QPLIB Benchmark (non-binary) (20 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 97
⭐ virtual best 0.25 95%
🥇 Gurobi-10.0.0 1.00 (1.00) 84%
🥈 OCTERACT-4.5.1 2.50 (5.65) 76%
🥉 SCIP-8.0.0 8.25 (18.70) 38%
📊 Baron-22.9.1 13.23 (30.00) 32%
📊 ANTIGONE-1.1 14.14 (32.00) 30%
📊 MINOTAUR-0.3.0 17.15 (38.80) 15%
📊 COUENNE-0.5 26.73 (60.50) 8%
previous benchmarks 🔽

Continuous Non-Convex QPLIB Benchmark (7 Jan 2023)

Choose base solver for comparison:

solver score (as reported) solved of 69
⭐ virtual best 0.22 91%
🥇 GUROBI-10.0.0 1.00 (1.00) 62%
🥈 OCTERACT-4.5.1 1.33 (1.72) 65%
🥉 ANTIGONE-1.1 3.85 (4.99) 41%
📊 Baron-22.9.1 4.51 (5.85) 28%
📊 MINOTAUR-0.3.0 5.43 (7.04) 22%
📊 SCIP-8.0.0 7.65 (9.92) 19%
📊 COUENNE-0.5 8.69 (11.30) 12%
📊 RAPOSa-3.2.0 10.96 (14.20) 6%
previous benchmarks 🔽

Convex Continuous QPLIB Benchmark (26 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 32
⭐ virtual best 0.85 100%
🥇 COPT-6.0.2 1.00 (1.00) 100%
🥈 MOSEK-10.0.18 1.33 (1.33) 100%
🥉 KNITRO-13.0.0 1.53 (1.53) 100%
📊 Gurobi-10.0.0 1.95 (1.95) 94%
📊 IPOPT-3.14.5 7.29 (7.29) 91%
previous benchmarks 🔽

Convex Discrete QPLIB Benchmark (12 Nov 2022)

Choose base solver for comparison:

solver score (as reported) solved of 31
⭐ virtual best 0.70 87%
🥇 GUROBI-10.0.0 1.00 (1.00) 81%
🥈 COPT-6.0.1 1.09 (1.09) 77%
🥉 OCTACT 6.72 (6.72) 68%
📊 Shot-1.1 8.42 (8.42) 48%
📊 MOSEK-10.0.18 8.96 (8.96) 61%
📊 KNITRO-13.1.0 12.88 (12.90) 52%
📊 Baron-22.9.1 13.61 (13.60) 58%
📊 SCIP-8.0.0 35.52 (35.50) 39%
📊 MNTAUR 40.70 (40.70) 45%
📊 Bonmin-1.8.7 52.55 (52.60) 23%
previous benchmarks 🔽

Mixed Integer Nonlinear Programming Benchmark (MINLPLIB) (29 Jan 2023)

Choose base solver for comparison:

solver score (as reported) solved of 87
⭐ virtual best 0.35 100%
🥇 OCTERACT 1.00 (0.72) 99%
🥈 BARON 1.39 (1.00) 84%
🥉 SCIP 2.77 (1.94) 74%
📊 ANTIGONE 10.42 (7.50) 61%
📊 LINDO 13.71 (9.87) 43%
📊 COUENNE 23.58 (17.00) 28%
previous benchmarks 🔽

MPEC Benchmark (Math. Progr. w. Equilibrium Constraints) (12 Apr 2022)

Choose base solver for comparison:

solver score (as reported) solved of 29
⭐ virtual best 1.00 83%
🥇 KNITRO-13.0 1.00 (1.00) 83%
🥈 filter-MPEC 7.82 (8.90) 62%
🥉 LOQO-7.03 16.69 (19.50) 21%