> ## Documentation Index
> Fetch the complete documentation index at: https://docs.compose.market/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Catalog

> 700+ models from multiple families and modalities through a single x402-native API.

Compose.Market provides unified access to AI models through a single API with x402-native payment. The curated catalog has 700+ models across multiple families and modalities.

## Two catalogs

| Catalog     | Access                                                      |
| ----------- | ----------------------------------------------------------- |
| **Curated** | `GET /v1/models` — verified pricing, validated capabilities |

## Providers

| Provider   | Specialization                                                   |
| ---------- | ---------------------------------------------------------------- |
| Alibaba    | Text (Qwen), speech (CosyVoice), video (Wan), vision, embeddings |
| Roboflow   | Object detection, segmentation, OCR, image embeddings            |
| Cloudflare | Edge inference, LLMs, vision, speech via Workers AI              |
| Gemini     | Multimodal text, vision, realtime, TTS, video                    |
| Azure      | Hosted OpenAI models with data residency                         |
| OpenAI     | GPT, Sora, embeddings, realtime                                  |
| Fireworks  | Open-source LLMs and rerankers                                   |
| Cartesia   | Low-latency text-to-speech                                       |
| ElevenLabs | Voice generation, music, realtime transcription                  |
| Deepgram   | Speech recognition (Nova-3)                                      |
| ASICloud   | Specialized inference                                            |

## Modalities

* Text generation
* Speech
* Image generation
* Video generation
* Realtime
* Embeddings

Specialized vision operations — detection, segmentation, OCR, keypoints, classification — are exposed through Roboflow models in the image/vision modality.

## Model examples

| Model                    | Provider          | Modality               |
| ------------------------ | ----------------- | ---------------------- |
| `gpt-5.5`                | Azure / OpenAI    | Text                   |
| `FLUX.2-pro`             | Black Forest Labs | Image                  |
| `sora-2`                 | OpenAI            | Video                  |
| `cosyvoice-v3-flash`     | Alibaba           | Speech                 |
| `gemini-3-flash-preview` | Gemini            | Text, vision, realtime |
| `qwen3`                  | Alibaba           | Text                   |
| `deepseek-v4-flash`      | DeepSeek          | Text                   |
| `text-embedding-3-large` | OpenAI            | Embeddings             |

## Quick example

```bash theme={null}
curl https://api.compose.market/v1/chat/completions \
  -H "Authorization: Bearer compose-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "messages": [{"role": "user", "content": "Hello"}]
  }'
```

## API compatibility

All models use OpenAI-compatible endpoints. The model string is the contract — Compose resolves the provider from the catalog row, not by inspecting the prompt. Duplicate upstream model IDs get explicit public aliases (e.g. `gpt-5.5` → Azure, `openai/gpt-5.5` → OpenAI).

## Pricing

The API applies a **1% platform fee** on top of the raw provider amount. OpenRouter charges 5.5%. Compose charges 1% — a 5.5x difference that is factual. Pricing is dynamic: each model's price comes from the generated catalog, not a fixed table.

## Related

* [Inference](/inference/introduction) — full API surface, native and external routes
* [SDK Inference](/sdk/inference/overview) — TypeScript API for model calls
* [Quickstart](/inference/quickstart) — list models, chat, and stream
* [Metering](/inference/metering) — how provider output becomes a bill