Introduction

Model Graph is a public API and canonical database that tracks AI models across every major provider and open-source ecosystem. It catalogs model families, individual versions, release and deprecation timelines, parameter counts, context windows, and the many aliases each model is known by across different SDKs and platforms.

Think of it as a package registry — but for AI models. Instead of tracking library versions, it tracks model versions: when they were released, when they're being deprecated, what replaces them, and every string identifier used to reference them in code.

The Problem

AI model identifiers are fragmented and chaotic:

• The same model has different IDs across providers, cloud platforms, and SDKs — claude-3-5-sonnet-20241022 vs. anthropic.claude-3-5-sonnet-20241022-v2:0 vs. anthropic/claude-3-5-sonnet-20241022
• New versions ship constantly, often with subtle naming changes
• Deprecations happen on different schedules per provider, announced in blog posts and changelog pages with no unified API
• Open-source models have parameter variants, fine-tuned derivatives, and no standard naming convention
• There is no single source of truth that unifies all of this

The Solution

Model Graph aggregates model metadata from every major source into one normalized, queryable database — kept current by automated ingestion pipelines that poll provider APIs, deprecation feeds, and Hugging Face daily.

What It Tracks

Dimension	Examples
Providers	Anthropic, OpenAI, Google, Meta, Mistral, Cohere, xAI
Model families	Claude Sonnet, GPT-4o, Llama 3, Gemini Pro
Model versions	Individual releases with full lifecycle metadata
Parameters	Parameter counts and labels (8B, 70B, 405B)
Context windows	Input and output token limits
Aliases	Every known string representation across SDKs and platforms
Upgrade paths	Which model replaces which
Canonical URLs	Links to Hugging Face pages for open-source models

Who It's For

• Developer tools that need model intelligence — like AI Updater Bot, which uses this registry to detect outdated model references in code and suggest upgrades
• Platform teams building internal AI tooling who need to know "what's the latest Sonnet?" or "is this model deprecated?" programmatically
• Developers who want a single API to query model capabilities, compare versions, or check deprecation status
• Open-source projects that need a canonical model database without building their own scraping infrastructure

Quick Example

Find out if a model string is outdated and get its recommended upgrade:

curl https://api.modelgraph.ai/api/v1/resolve?q=claude-3-sonnet-20240229

{
  "matched_model": {
    "slug": "claude-3-sonnet-20240229",
    "display_name": "Claude 3 Sonnet",
    "status": "deprecated",
    "family": "claude-sonnet"
  },
  "matched_alias": {
    "alias": "claude-3-sonnet-20240229",
    "source": "official"
  },
  "confidence": 1.0,
  "is_latest": false,
  "upgrade": {
    "slug": "claude-sonnet-4-6",
    "display_name": "Claude Sonnet 4.6",
    "status": "latest"
  },
  "upgrade_alias": {
    "alias": "claude-sonnet-4-6",
    "source": "official"
  }
}

Next Steps

• Quickstart — Make your first API call in 60 seconds
• API Reference — Full endpoint documentation
• Concepts — Understand model lifecycles, aliases, and data freshness
• Examples — Code samples in Python, JavaScript, Go, and curl