Skip to main content
Compose Market provides unified access to AI models from multiple providers through a single API.

​
Model Catalog

The platform maintains two catalogs:

​
Curated Catalog (~2,100 models)

Models visible on the platform with:
  • Verified pricing
  • Validated capabilities
  • Quality assurance
Access via: GET /v1/models

​
Extended Catalog (50,000+ models)

Complete index from all providers including:
  • Experimental models
  • Community fine-tunes
  • Specialized variants
Access via: GET /v1/models/all

​
Providers

ProviderModelsSpecialization
OpenAIGPT-4o, o1, DALL-EGeneral, reasoning
AnthropicClaude 4, 3.7Analysis, coding
GoogleGemini 3, Veo, LyriaMultimodal
AIMLVariousAggregation
HuggingFaceOpen sourceCommunity models
OpenRouterRouting200+ models

​
Model Types

​
Text Generation

Models for chat, reasoning, and content generation:
  • google/gemini-3-flash-preview - Fast multimodal
  • deepseek/deepseek-v3.2 - Reasoning
  • arcee-ai/trinity-mini - Efficient

​
Image Generation

Text-to-image and image-to-image:
  • fal-ai/flux-2 - High quality
  • bytedance/seedream-4-5 - Fast generation
  • dall-e-3 - OpenAI

​
Video Generation

Text-to-video and image-to-video:
  • openai/sora-2 - Cinematic
  • google/veo3-1 - Fast
  • kwai/kling-2.6 - Motion control

​
Embeddings

Feature extraction for search and similarity:
  • text-embedding-3-large - OpenAI
  • intfloat/e5-mistral-7b-instruct - Open source

​
Audio

Speech and music generation:
  • tts-1-hd - Text-to-speech
  • whisper-1 - Speech recognition
  • google/lyria2 - Music generation

​
ModelCard Schema

Each model has a ModelCard with: 
interface ModelCard {
  modelId: string;           // Canonical identifier
  name: string;              // Display name
  provider: string;          // Provider ID
  taskType: string;          // Primary task
  capabilities: string[];    // Supported features
  pricing: {
    inputTokenPrice: string;
    outputTokenPrice: string;
  };
  contextWindow?: number;
  maxOutputTokens?: number;
}

​
API Compatibility

All models use OpenAI-compatible endpoints:
  • POST /v1/chat/completions - Text generation
  • POST /v1/images/generations - Image generation
  • POST /v1/embeddings - Embeddings
This allows drop-in replacement with existing OpenAI client libraries.