model-metadata-central 0.4.0

Typed registry of LLM model metadata for Python — context windows, pricing, provider routing.

model-metadata-central

Typed registry of LLM model metadata for Python — context windows, pricing, provider routing.

pip install model-metadata-central

Usage

Full registry

from model_metadata_central import get_all_models, get_all_providers

models = get_all_models()
for model in models:
    print(f"{model['model_name']}: {model['context_window']} context")

Look up a single model

from model_metadata_central import get_model

model = get_model("gpt-4o")
if model:
    print(model["context_window"])          # 128000
    print(model["cost_per_million_tokens"]) # {"input": 5, "output": 15}

Filter by provider

from model_metadata_central import get_models_by_provider

anthropic = get_models_by_provider("anthropic")
openrouter = get_models_by_provider("openrouter")

Provider routing

from model_metadata_central import get_provider_model_id, get_provider

# Get the model ID for a specific provider
openai_id = get_provider_model_id("gpt-4o", "openai")           # "gpt-4o"
orouter_id = get_provider_model_id("gpt-4o", "openrouter")     # "openai/gpt-4o"

# Get provider config
provider = get_provider("openai")
print(provider["base_url"])    # "https://api.openai.com/v1"
print(provider["auth_type"])    # "api_key"

Named model constants

from model_metadata_central import GPT_4_O, CLAUDE_OPUS_4_7

# Fully typed dict — bundlers exclude the rest
print(GPT_4_O["context_window"])

API

Function	Returns	Description
get_model(id)	dict | None	Lookup by model_id
get_all_models()	list[dict]	All models
get_models_by_provider(provider_id)	list[dict]	Filter by provider
get_model_on_provider(provider_id, provider_model_id=None)	dict | None	Find model on a specific provider
get_provider(id)	dict | None	Provider config
get_all_providers()	list[dict]	All providers
get_provider_model_id(model_id, provider_id)	str | None	Provider-specific model ID

Data

• Registry pre-compiled from models/*.yaml at install/build time
• 73 models across 18 providers
• Prices are in USD per 1,000,000 tokens

Schema

Matches the JSON Schema definitions in the repo root.