python data models performance

Overview

three approaches to data modeling in python:

• dataclasses: standard library data containers
• pydantic: validation and serialization
• sqlalchemy: database orm

this benchmark measures their relative performance.

Test Setup

each library implements the same model:

# Dataclass
@dataclass
class PersonDataclass:
    name: str
    age: int
    email: str
    active: bool = True

# Pydantic
class PersonPydantic(BaseModel):
    name: str
    age: int
    email: str
    active: bool = True

# SQLAlchemy (with in-memory SQLite)
class PersonSQLAlchemy(Base):
    __tablename__ = 'persons'
    id = Column(Integer, primary_key=True)
    name = Column(String)
    age = Column(Integer)
    email = Column(String)
    active = Column(Boolean, default=True)

Operations Tested

• creation: instantiating objects
• modification: updating attributes
• serialization: converting to dictionaries

10,000 iterations per operation, measured with time.perf_counter()

Results

Average Performance (Python 3.10-3.14)

Key Findings

• pydantic serializes ~2.5x faster due to rust-based implementation
• sqlalchemy operates at a different scale (50-140x slower) due to database operations
• dataclasses and pydantic have similar performance for basic operations

Actual Times (Microseconds)

When to Use Each

Dataclasses

• simple data containers
• validation not required
• zero dependencies
• significant performance improvements in python 3.13+

Pydantic

• data validation required
• json serialization workflows
• api development (fastapi)
• configuration management

SQLAlchemy

• database persistence
• complex queries and relationships
• transaction management
• database-agnostic applications

Benchmark Files

• benchmark.py - benchmark runner
• analyze.py - data analysis
• visualize.py - visualization
• summary.json - aggregated results
• results_python_*.json - raw data per python version

Running the Benchmarks

# Run benchmarks with specific Python version
uv run --python 3.14 --prerelease allow --with pydantic --with sqlalchemy python benchmark.py

# Analyze results
uv run --with matplotlib python analyze.py

# Generate visualizations
uv run --with matplotlib python visualize.py

Conclusion

performance is one factor among many:

• dataclasses: optimal for simple use cases, especially with python 3.13+
• pydantic: best choice for validation and serialization workflows
• sqlalchemy: necessary for database-backed applications

choose based on requirements, not benchmarks alone.

See Also

• performance across python versions
• choosing a data model
• dataclasses documentation
• pydantic documentation
• sqlalchemy documentation