Harrington Park, NJ — January 1st, 2025 — We have achieved a new best feasible solution for the MIPLIB benchmark instance bharat, improving the incumbent objective from 4,170,993 to 4,139,915.438 using our mixed-integer optimization workflow.

This improvement represents a 31,077.562 reduction in the objective (approximately 0.75%), strengthening the best-known solution quality for this challenging benchmark model. The run concluded with a remaining 8.25% optimality gap at termination.

“This result reflects our continued progress on difficult benchmark instances—where incremental improvements require both strong modeling discipline and robust solver performance,” said Robert Ashford, CEO.

Result Summary

• Benchmark library: MIPLIB
• Instance: bharat
• Link: https://miplib.zib.de/instance_details_bharat.html
  
• Previous incumbent (objective): 4,170,993
• New best feasible solution (objective): 4,139,915.438
• Improvement: 31,077.562 (~0.75%)
• Gap at termination: 8.25%
• Run metrics (this solve): runtime 2,752.68; search metric 2,746 (from our benchmark log)

About MIPLIB

MIPLIB is a widely used public benchmark library for mixed-integer programming, supporting transparent and repeatable comparisons of optimization performance across diverse, real-world model types.