Gurobi Reporting Incorrect Optimal Solution

Answered

Kathy Wang

September 03, 2025 15:44

• 
• 
• 

I have been noticing some issues lately (in the past 2 weeks) where both Gurobi 11.0.2 and 12.0.3 have returned suboptimal solutions with a 0% optimality gap. In the log file, it appears that the lower bound suddenly jumps to be exactly the best integer solution so far, and terminates the run. Strangely enough, I have never encountered these issues before, despite working with the same models and instances for several months now.

For example, here is a snippet of the log file under Gurobi version 11.02, where I have found a better solution of 128,142 using the exact same model:

Similarly, in Gurobi 12.0.3, where the same model has found a better solution of 128,344 in a previous run using Gurobi 11.0.2 (which has not been proven optimal yet)

0

• Official comment

  

  Nigora Gafur

  • Gurobi Staff

  September 08, 2025 10:27

  Hi Kathy,

  Based on the provided log files, I cannot confirm that you are using exactly the same models, as the log file obtained with Version 12 is incomplete. However, Gurobi found an optimal solution for both models with a 0% optimality gap. Two models are considered identical if their model fingerprints are the same, regardless of the Gurobi version used. For more information, please see: What is the model fingerprint?.

  I can see that the model solved with Version 11 was provided with a MIP start. I cannot confirm whether the same was done for the model in Version 12. Please note that using different MIP start values can lead to different model fingerprints.

  If you're looking to compare the performance of different Gurobi versions, make sure to use the exact same models and provide the same MIP start values.

  I hope this helps answer your question. Don’t hesitate to reach out if you have any further questions.

  Best

  Nigora

Please sign in to leave a comment.