faker-ai-provider 2.2.0

Generate realistic AI/ML test data: 180+ models, 30+ companies, 20+ frameworks, 30+ datasets, architectures, tasks, and parameters

faker-ai-provider

Faker provider for generating AI/ML-related fake data with correlated relationships between models, companies, architectures, and capabilities.

Model Data Updated: July 2026 (includes Claude Fable 5, GPT-5.5, Gemini 3.5 Flash, DeepSeek-V4, Qwen3.7-Max, Grok 4.3, Kimi K2.6, and more)

Installation

pip install faker-ai-provider

Quick Start

from faker import Faker
from faker_ai import AiProvider

fake = Faker()
fake.add_provider(AiProvider)

# Generate correlated AI data
fake.ai_model()           # 'Claude Opus 4.7'
fake.ai_company()         # 'Anthropic'
fake.full_ai_model_spec() # 'gpt-oss-120b by OpenAI: Transformer architecture, 120B parameters, for reasoning.'

Seeding for Reproducibility

Use Faker's seeding to generate consistent, reproducible data across runs:

fake = Faker()
fake.add_provider(AiProvider)
fake.seed_instance(42)

# These will always return the same values with seed 42
print(fake.ai_model())    # Always 'LLaMA 3 70B'
print(fake.ai_company())  # Always 'Apple'

Available Methods

Basic Methods

Method	Example
ai_model()	GPT-5.3-Codex, Claude Opus 4.7, Gemini 3 Pro Preview
ai_company()	OpenAI, Anthropic, Google DeepMind
ai_architecture()	Transformer, Diffusion, Mixture of Experts
ai_task()	text-generation, code-generation, reasoning
ai_modality()	text, image, audio, video
ml_framework()	PyTorch, TensorFlow, LangChain
ai_dataset()	ImageNet, COCO, MMLU, FineWeb

Correlation Methods

Method	Description
ai_model_for_company(company)	Get a model from a specific company
ai_company_for_model(model)	Get the company that created a model
ai_tasks_for_model(model)	Get tasks supported by a model
ai_models_for_task(task)	Get models that support a task
ai_models_by_architecture(arch)	Filter models by architecture
ai_models_by_modality(modality)	Filter models by modality
model_scenario(model=None)	Get complete correlated model data

Composite Methods

Method	Description
full_ai_model_spec()	Formatted spec: "Model by Company: arch, params, for task."
ai_training_run()	Dict with model, framework, dataset, task
ai_deployment()	Dict with model, endpoint, version, status
ai_experiment()	Dict with experiment_id, accuracy, loss, epochs

Advanced Usage

Populate a Database with AI Records

from faker import Faker
from faker_ai import AiProvider

fake = Faker()
fake.add_provider(AiProvider)

# Generate 100 AI deployment records
deployments = [fake.ai_deployment() for _ in range(100)]

# Generate experiment tracking data
experiments = [fake.ai_experiment() for _ in range(50)]

Generate ML Pipeline Configuration

fake.seed_instance(42)  # Reproducible pipeline

pipeline = {
    "name": f"pipeline-{fake.random_int(1000, 9999)}",
    "training": fake.ai_training_run(),
    "deployment": fake.ai_deployment(),
    "experiment": fake.ai_experiment(),
}

Filter Models by Capability

# Get all models that support code generation
code_models = fake.ai_models_for_task("code-generation")

# Get all diffusion models
diffusion_models = fake.ai_models_by_architecture("Diffusion")

# Get all multimodal models
video_models = fake.ai_models_by_modality("video")

Model Scenario

Get complete, correlated model information:

scenario = fake.model_scenario()
# {
#     'model': 'GPT-5.2',
#     'company': 'OpenAI',
#     'architecture': 'Transformer',
#     'modality': ['text', 'image', 'audio', 'video'],
#     'tasks': ['text-generation', 'reasoning', 'code-generation', ...],
#     'parameters': 'undisclosed',
#     'release_year': 2026
# }

License

MIT