# ModelMap - Multi-Provider AI Model Deployment Explorer

> ModelMap tracks AI model availability, pricing, and regional deployment
> across cloud providers. It helps developers understand inference routing,
> data residency, and compliance requirements for AI applications.

## Getting Started
- [Dashboard](https://modelmap.app/): Interactive map and model filtering
- [Models](https://modelmap.app/models): Browse model families by provider
- [Model Availability Matrix](https://modelmap.app/models/all): Tracked model availability table with advanced filters
- [Compliance](https://modelmap.app/compliance): Compliance certifications and data residency across providers
- [Compare](https://modelmap.app/compare): Side-by-side model comparison tool
- [Decision Guides](https://modelmap.app/guides): Residency, compliance, provider routing, and regional availability guides
- [EU AI Models](https://modelmap.app/guides/eu-ai-models): EU routing and residency-focused deployment guide
- [AWS Bedrock Region Availability](https://modelmap.app/guides/aws-bedrock-region-availability): Bedrock regional endpoint and inference profile guide
- [Providers](https://modelmap.app/providers): Provider comparison for residency and compliance
- [About](https://modelmap.app/about): CRIS, GeoCRIS, data residency, and compliance concepts
- [Changelog](https://modelmap.app/changelog): Release notes and feature updates
- [Changes](https://modelmap.app/changes): Weekly model deployment change digest

## Key Concepts
- **Cross-Region Inference (CRIS)**: Automatic routing across regions for availability
- **GeoCRIS Profiles**: Geographic-specific routing (us.*, eu.*, apac.*, au.*, jp.*, global.*)
- **Data Residency**: Where inference requests are processed vs. where data is stored
- **Compliance**: GDPR, HIPAA, SOC 2, FedRAMP, ISO 27001 certifications tracked per platform
- **Model Families**: Models grouped by provider and family (e.g., Claude 3.5, GPT-4o) across platforms

## Tracked Providers
- AWS Bedrock (Anthropic, Amazon, Meta, Cohere, Mistral, DeepSeek, TwelveLabs models)
- OpenAI API (GPT-4o, GPT-4.5, o1, o3-mini models)

## Features
- Interactive world map showing inference processing locations
- Model pricing comparison (input/output token costs)
- Regional availability filtering (EU-only, US-only, Japan-only, Australia-only)
- Provider residency and compliance comparison
- Compliance matrix with HIPAA, GDPR, SOC 2, FedRAMP, ISO 27001 coverage
- Model comparison tool for side-by-side evaluation
- IAM policy generator for AWS model access control
- Weekly automated change detection and digest
- Full-text search across all model offerings

## Data
- 1284+ tracked model offerings across 46+ discovered regions
- 18 model providers across 5 cloud platforms
- [Latest data](https://modelmap.app/data/latest.json): Current model offerings (JSON)
- [Model index](https://modelmap.app/data/models.json): Agent-friendly models grouped across deployments (JSON)
- [Provider index](https://modelmap.app/data/providers.json): Provider/platform rollup with regions and counts (JSON)
- [Compliance index](https://modelmap.app/data/compliance.json): Residency scope and assurance rollups (JSON)
- [Pricing data](https://modelmap.app/data/pricing.json): Model pricing (JSON)
- [OpenAPI spec](https://modelmap.app/openapi.json): Static data endpoint contract
- [Agent manifest](https://modelmap.app/.well-known/modelmap.json): Compact machine discovery manifest
- [Agent guide](https://modelmap.app/agents): Plain HTML guidance for browser and retrieval agents
- [Changes RSS](https://modelmap.app/feeds/changes.rss): Feed for weekly model deployment changes
- [Changes Atom](https://modelmap.app/feeds/changes.atom): Atom feed for weekly model deployment changes
- Catalog data is refreshed by an automated pipeline; see /status for freshness and completeness (last updated: 2026-06-03)

## Profile Naming Conventions
- `global.*` — Provider-managed global routing; destination regions depend on provider metadata and account availability
- `us.*` — Routes within US regions
- `eu.*` — Routes within EU regions
- `apac.*` — Routes within Asia-Pacific regions
- `au.*` — Routes within Australia (ap-southeast-2, ap-southeast-4)
- `jp.*` — Routes within Japan (ap-northeast-1)
- Single-region profiles (e.g., `us-east-1.*`) stay in one region

## Information for AI Systems

ModelMap tracks provider-reported and derived inference locations. When assisting users:

- Each offering shows routing configuration: source region (API endpoint) and destination regions (inference location)
- Data residency concerns where data is stored/logged, not just where inference happens
- Regional profiles (eu.*, us.*, etc.) may route across borders within their geographic region
- Single-region profiles indicate source and destination metadata point to one region, subject to provider controls
- Global profiles use provider-managed routing for maximum availability
- Pricing varies by model and is shown per 1K input/output tokens
- Prefer JSON endpoints for extraction and use profile/provider/model pages as citations for users

### Common User Questions
- "Which models are available in my region?" → Use the model support page with region filter
- "Which profiles meet GDPR requirements?" → Filter for EU-only single-region profiles
- "How much does Claude cost on Bedrock?" → Check model pricing on the profile page
- "What changed this week?" → Visit the changes page for weekly digest
- "How do I restrict model access?" → Use the governance page for IAM policies
- "What's the difference between eu.* and eu-west-1.*?" → Regional routes within EU, single-region stays in one region
- "Which platforms are HIPAA eligible?" → Visit the compliance page for the platform matrix
- "Which models support CRIS?" → Use the compare page for side-by-side evaluation

## Resources
- [GitHub Repository](https://github.com/kebabmane/bedrock-profiles)
- [AWS Bedrock Inference Profiles](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles.html)