# AI Model Compatibility Explorer

> Which LLMs take images, audio or files, fit your context and support tools or structured outputs? Filters and compares hundreds of language models by input modalities (image, audio, file, video), output modality, context window, maximum output tokens and API features such as tools, tool_choice, structured_outputs, response_format and reasoning, using OpenRouter's public models API (or a pasted copy of it). Shows matching models with prices, near misses with what they lack, and recommendations.

Part of usefulweb.site (https://usefulweb.site/). Results are deterministic calculations from the inputs and the listed assumptions; the API returns exactly what the web page shows.

## Use it
- Human interface: https://aimodelcompatibilityexplorer.usefulweb.site/ (interactive tool at https://aimodelcompatibilityexplorer.usefulweb.site/tool)
- JSON API: POST https://aimodelcompatibilityexplorer.usefulweb.site/api/v1/calculate with a JSON object of inputs (schema: https://aimodelcompatibilityexplorer.usefulweb.site/api/v1/tool)
- OpenAPI: https://aimodelcompatibilityexplorer.usefulweb.site/openapi.json
- MCP (Streamable HTTP): https://aimodelcompatibilityexplorer.usefulweb.site/mcp - tool `ai_model_compatibility_explorer`

## Inputs
- source (select): Model data
- modelsJson (textarea): Models JSON - Response of GET https://openrouter.ai/api/v1/models (or a subset).
- search (text): Name / provider contains
- inputModalities (multiselect): Required input types (besides text)
- outputModality (select): Required output
- minContext (integer): Minimum context window
- minOutputTokens (integer): Minimum max output
- parameters (multiselect): Required API features
- sort (select): Sort by
- maxResults (integer): Maximum models shown

## Method
- Live mode reads GET https://openrouter.ai/api/v1/models (public, no key). Fields used: architecture.input_modalities/output_modalities, context_length, top_provider.max_completion_tokens, supported_parameters and pricing (USD per token, shown per 1M tokens).
- A model matches when it has every required input modality, the required output modality, at least the minimum context and output tokens, and every required API parameter listed in supported_parameters.
- Near misses are models (matching the name filter) that lack exactly one requirement.
- Recommendations: cheapest by input price, largest context and broadest feature set among the matches. Prices are for the routed provider listed by OpenRouter, not necessarily the vendor's own API.