Models

Overview

A model is a satisfying assignment for a set of formulas. When Z3 determines a formula is satisfiable, it can produce a model that shows concrete values making the formula true.

Getting Models

From examples/c++/example.cpp:45:

from z3 import *

x = Int('x')
y = Int('y')

s = Solver()
s.add(x >= 1)
s.add(y < x + 3)

if s.check() == sat:
    m = s.model()
    print(m)
    # Model:
    #   x = 1
    #   y = 0

Model Evaluation

From examples/c++/example.cpp:56:

from z3 import *

x = Int('x')
y = Int('y')

s = Solver()
s.add(x >= 1)
s.add(y < x + 3)

if s.check() == sat:
    m = s.model()
    
    # Evaluate individual variables
    print(f"x = {m[x]}")  # x = 1
    print(f"y = {m[y]}")  # y = 0
    
    # Evaluate expressions
    print(f"x + y + 1 = {m.eval(x + y + 1)}")  # 2
    print(f"x * y = {m.eval(x * y)}")          # 0

Iterating Over Models

From examples/c++/example.cpp:58:

from z3 import *

x = Int('x')
y = Int('y')

s = Solver()
s.add(x >= 1)
s.add(y < x + 3)

if s.check() == sat:
    m = s.model()
    
    # Iterate over all declarations in model
    for decl in m.decls():
        print(f"{decl.name()} = {m[decl]}")

Structured Iteration

From examples/c++/example.cpp:59:

from z3 import *

x = Int('x')
y = Int('y')

s = Solver()
s.add(x >= 1)
s.add(y < x + 3)

if s.check() == sat:
    m = s.model()
    
    # Model has size (number of interpretations)
    print(f"Model size: {len(m)}")
    
    for i in range(len(m)):
        decl = m[i]  # Get i-th declaration
        
        # For constants (0-arity functions)
        if decl.arity() == 0:
            print(f"{decl.name()} = {m[decl]}")
        else:
            # For functions, see function interpretations below
            print(f"{decl.name()} is a function")

Model Completeness

Models may not assign values to all variables in your problem—only those relevant to satisfying the constraints.

from z3 import *

x = Int('x')
y = Int('y')
z = Int('z')  # Unconstrained

s = Solver()
s.add(x > y)

if s.check() == sat:
    m = s.model()
    
    print(f"x = {m[x]}")  # Has value
    print(f"y = {m[y]}")  # Has value
    print(f"z = {m[z]}")  # May be None or default value
    
    # Use eval with model_completion for unconstrained vars
    print(m.eval(z, model_completion=True))  # Returns arbitrary value

Function Interpretations

Models can include interpretations of uninterpreted functions: From examples/c++/example.cpp:77:

from z3 import *

x, y = Ints('x y')
f = Function('f', IntSort(), IntSort())

s = Solver()
s.add(x == y)
s.add(f(x) != f(y))

if s.check() == sat:
    m = s.model()
    print(m)
else:
    print("Unsatisfiable")  # This will be unsat

Inspecting Function Interpretations

from z3 import *

x, y = Ints('x y')
f = Function('f', IntSort(), IntSort())

s = Solver()
s.add(f(0) == 10)
s.add(f(1) == 20)
s.add(f(2) == 30)

if s.check() == sat:
    m = s.model()
    
    # Get function interpretation
    f_interp = m[f]
    
    print("Function f interpretation:")
    print(f_interp)
    
    # Evaluate function at specific points
    print(f"f(0) = {f_interp(0)}")  # 10
    print(f"f(1) = {f_interp(1)}")  # 20
    print(f"f(5) = {f_interp(5)}")  # Default value (else case)

Array Models

from z3 import *

A = Array('A', IntSort(), IntSort())
i, j = Ints('i j')

s = Solver()
s.add(A[i] == 10)
s.add(A[j] == 20)
s.add(i != j)

if s.check() == sat:
    m = s.model()
    
    print("Array A interpretation:")
    print(m[A])
    
    # Evaluate at specific indices
    print(f"A[{m[i]}] = {m.eval(A[m[i]])}")
    print(f"A[{m[j]}] = {m.eval(A[m[j]])}")

Model Evaluation Options

from z3 import *

x = Int('x')
y = Int('y')

s = Solver()
s.add(x > 0)

if s.check() == sat:
    m = s.model()
    
    # Default evaluation
    val1 = m.eval(x + y)
    
    # Force model completion (assign values to unconstrained vars)
    val2 = m.eval(x + y, model_completion=True)
    
    print(f"Without completion: {val1}")
    print(f"With completion: {val2}")

Creating Models Programmatically

From examples/c++/example.cpp:1354:

from z3 import *

# Create empty model (requires Z3 4.8.8+)
ctx = Context()
m = Model(ctx)

# Add constant interpretations
a = Int('a')
b = Int('b')

m.add_const_interp(a.decl(), 0)
m.add_const_interp(b.decl(), 1)

# Evaluate in constructed model
result = m.eval(a + b < 2)
print(result)  # True

Model Conversion

When using tactics, you may need to convert models between goals: From examples/c++/example.cpp:680:

from z3 import *

x, y, z = Ints('x y z')

# Create and transform goal
g = Goal()
g.add(x > 10)
g.add(y == x + 3)
g.add(z > y)

t = Then('simplify', 'solve-eqs')
result = t(g)

# Solve transformed subgoal
s = Solver()
subgoal = result[0]
for f in subgoal:
    s.add(f)

if s.check() == sat:
    # Model for transformed problem
    m_sub = s.model()
    print("Subgoal model:", m_sub)
    
    # Convert back to original variables
    m_orig = subgoal.convert_model(m_sub)
    print("Original model:", m_orig)

Model String Representation

from z3 import *

x, y = Ints('x y')
s = Solver()
s.add(x > y)

if s.check() == sat:
    m = s.model()
    
    # Default representation
    print(str(m))
    
    # SMT-LIB format
    print(m.sexpr())

Evaluating Complex Expressions

from z3 import *

x, y, z = Ints('x y z')
f = Function('f', IntSort(), IntSort())

s = Solver()
s.add(x == 5)
s.add(y == x + 2)
s.add(f(x) == 10)

if s.check() == sat:
    m = s.model()
    
    # Evaluate nested expressions
    print(m.eval(f(x) + y))        # 17
    print(m.eval(f(f(x))))         # f(10)
    
    # Evaluate Boolean formulas
    print(m.eval(x > 0))           # True
    print(m.eval(And(x > 0, y < 0))) # False

Datatype Models

from z3 import *

# Define list datatype
List = Datatype('List')
List.declare('nil')
List.declare('cons', ('head', IntSort()), ('tail', List))
ListSort = List.create()

# Constructors
nil = ListSort.nil
cons = ListSort.cons

# Accessors
head = ListSort.head
tail = ListSort.tail

l = Const('l', ListSort)

s = Solver()
s.add(l == cons(1, cons(2, cons(3, nil()))))

if s.check() == sat:
    m = s.model()
    print(m[l])
    # cons(1, cons(2, cons(3, nil)))
    
    print(m.eval(head(l)))        # 1
    print(m.eval(head(tail(l))))  # 2

Model Validation

Verify a model satisfies the original constraints:

from z3 import *

def validate_model(solver, model):
    """Check if model satisfies all assertions"""
    for assertion in solver.assertions():
        if not is_true(model.eval(assertion)):
            print(f"Failed: {assertion}")
            return False
    return True

x, y = Ints('x y')
s = Solver()
s.add(x > 0)
s.add(y > x)
s.add(x + y < 10)

if s.check() == sat:
    m = s.model()
    if validate_model(s, m):
        print("Model is valid!")

Enumerating Multiple Models

from z3 import *

def all_models(solver, vars, max_models=10):
    """Enumerate multiple models"""
    models = []
    count = 0
    
    while count < max_models and solver.check() == sat:
        m = solver.model()
        models.append(m)
        
        # Block this model
        block = Or([v() != m[v] for v in vars])
        solver.add(block)
        
        count += 1
    
    return models

x = Int('x')
y = Int('y')

s = Solver()
s.add(x >= 0, x <= 2)
s.add(y >= 0, y <= 2)

for i, m in enumerate(all_models(s, [x, y], max_models=5)):
    print(f"Model {i+1}: x={m[x]}, y={m[y]}")

Pretty Printing Models

from z3 import *

set_option("model.compact", False)  # Expanded format
set_option("model.v2", True)        # Version 2 format

x, y = Reals('x y')
s = Solver()
s.add(x*x + y*y == 1)
s.add(x > 0)

if s.check() == sat:
    m = s.model()
    print(m)
    
# Configure decimal precision
set_option("pp.decimal", True)
set_option("pp.decimal_precision", 10)

if s.check() == sat:
    m = s.model()
    print(m)

Model Parameters

from z3 import *

s = Solver()

# Configure model generation
s.set("model.compact", False)
s.set("model.v2", True)

x, y = Ints('x y')
s.add(x > y)

if s.check() == sat:
    m = s.model()
    print(m)

Best Practices

Check before accessing

Always verify check() == sat before calling model():

if s.check() == sat:
    m = s.model()  # Safe

Use model_completion wisely

model_completion=True assigns arbitrary values to unconstrained variables. Use when you need complete assignments.

Cache evaluations

Model evaluation can be expensive. Cache results if evaluating the same expression multiple times.

Validate models in debug mode

During development, validate that models actually satisfy constraints.

Solvers

Generating models with solvers

Expressions

Building expressions to evaluate

SMT Solving

Satisfiability and models

References

Model API: src/api/api_model.cpp
Model implementation: src/model/model.h
Examples: examples/c++/example.cpp:45-66, examples/c++/example.cpp:252-270

Getting Started

Core Concepts

Building Z3

Guides

Overview

Getting Models

Model Evaluation

Iterating Over Models

Structured Iteration

Model Completeness

Function Interpretations

Inspecting Function Interpretations

Array Models

Model Evaluation Options

Creating Models Programmatically

Model Conversion

Model String Representation

Evaluating Complex Expressions

Datatype Models

Model Validation

Enumerating Multiple Models

Pretty Printing Models

Model Parameters

Best Practices

Solvers

Expressions

SMT Solving

References

Build docs developers (and LLMs) love

Getting Started

Core Concepts

Building Z3

Guides

​Overview

​Getting Models

​Model Evaluation

​Iterating Over Models

​Structured Iteration

​Model Completeness

​Function Interpretations

​Inspecting Function Interpretations

​Array Models

​Model Evaluation Options

​Creating Models Programmatically

​Model Conversion

​Model String Representation

​Evaluating Complex Expressions

​Datatype Models

​Model Validation

​Enumerating Multiple Models

​Pretty Printing Models

​Model Parameters

​Best Practices

​Related Topics

Solvers

Expressions

SMT Solving

​References

Build docs developers (and LLMs) love

Overview

Getting Models

Model Evaluation

Iterating Over Models

Structured Iteration

Model Completeness

Function Interpretations

Inspecting Function Interpretations

Array Models

Model Evaluation Options

Creating Models Programmatically

Model Conversion

Model String Representation

Evaluating Complex Expressions

Datatype Models

Model Validation

Enumerating Multiple Models

Pretty Printing Models

Model Parameters

Best Practices

Related Topics

References