Microsoft Research Releases OptiMind: A 20B Parameter Model for Optimization

OptiMind: Natural Language to Optimization Models

Microsoft Research unveiled OptiMind, a 20B parameter Mixture of Experts model designed to convert natural language descriptions of optimization problems into executable mathematical formulations. This tackles a significant bottleneck in operations research where manual model creation often takes days of expert effort.

Why This Matters

Currently, translating real-world business problems into precise mathematical models (like Mixed Integer Linear Programs) requires specialized expertise. The complexity and potential for error in this process can lead to suboptimal solutions or even intractable problems, costing organizations significant time and resources – potentially millions in lost efficiency. OptiMind aims to automate this translation, reducing reliance on scarce expert resources and accelerating the problem-solving cycle.

Key Insights

• Model Size & Activation: OptiMind is a 20B parameter model, but only 3.6B parameters are active per token during inference, balancing capacity and cost.
• Class-Based Error Analysis: The model leverages a novel approach of classifying optimization problems into 53 seed classes to identify and correct common formulation errors during training.
• Azure AI Foundry Integration: OptiMind is available in Azure AI Foundry as microsoft-optimind-sft, enabling integration into existing decision support pipelines.

Working Example

# Example GurobiPy code generated by OptiMind (illustrative)
import gurobipy as gp
from gurobipy import GRB

try:
    # Create a new model
    model = gp.Model("supply_chain")

    # Create variables
    x = model.addVar(vtype=GRB.INTEGER, name="x")
    y = model.addVar(vtype=GRB.INTEGER, name="y")

    # Set objective function
    model.setObjective(x + 2*y, GRB.MAXIMIZE)

    # Add constraints
    model.addConstr(x + y <= 10, "c1")
    model.addConstr(x >= 0, "c2")
    model.addConstr(y >= 0, "c3")

    # Optimize the model
    model.optimize()

    # Print the optimal solution
    print('Optimal objective value:', model.objVal)
    print('x =', x.x)
    print('y =', y.x)

except gp.GurobiError as e:
    print('Error code ' + str(e.errno) + ': ' + str(e))

Practical Applications

• Supply Chain Optimization: A logistics company could use OptiMind to describe a distribution problem in natural language, and the model would generate the MILP to optimize delivery routes and warehouse assignments.
• Pitfall: Relying solely on the generated model without validation. Users should always verify the correctness of the generated formulation and constraints, as LLMs can still make logical errors.

References:

• https://www.marktechpost.com/2026/01/19/microsoft-research-releases-optimind-a-20b-parameter-model-that-turns-natural-language-into-solver-ready-optimization-models/