Models
Use this section to choose the right model family, request shape, and output modality before you hardcode a model ID.
Use this section before you hardcode a model ID or copy a provider example into production. Model choice in AnyInt includes three separate decisions: the compatibility family, the input/output modality, and the operational pattern for sync, streaming, or task-based work.
Choose by question
| Question | Page |
|---|---|
| Which model or compatibility family should I use? | Models |
| Which modalities does AnyInt expose? | Overview |
| How do I send image input to a model? | Images Input |
| How do I generate images? | Image Generation |
| How do I generate videos or poll video tasks? | Video Generation |
Recommended selection workflow
| Step | Decision | Where to confirm |
|---|---|---|
| 1 | Which account and environment will run this workload? | Your deployment configuration and AnyInt dashboard |
| 2 | Which model IDs are actually available? | Models API |
| 3 | Which compatibility family does your client already speak? | API Reference |
| 4 | Does the request need text, image, video, or music output? | Overview |
| 5 | Is the response sync, streaming, or task-based? | The API page for the selected route |
| 6 | What fallback should run if the primary path fails? | Model Fallbacks |
Route family quick map
| Workload | Start with | Notes |
|---|---|---|
| Text chat, summarization, extraction, or coding-agent requests | OpenAI-compatible API | Fastest first integration for most clients |
| Claude-style messages or image understanding with content blocks | Anthropic-compatible API | Use Anthropic headers and content block request shapes |
| Gemini-native generation, image models, or Gemini function declarations | Gemini-compatible API | Use contents[].parts[]; do not send OpenAI messages to Gemini routes |
| Prompt-driven image or video generation | Media API | Image responses can be sync; video is usually task-based |
| Songs, covers, lyrics, stems, or music videos | AI Music API | Treat creation as an async workflow with polling or callbacks |
Keep the API contract separate
This section helps you make product and workload choices. For the exact model discovery endpoint, use Models API. The current verified model list route is GET /openai/v1/models.
Do not infer support from a model name alone. A model can support text in one API family, image input in another, and task-based media generation through a separate route. The route page defines the request body customers can rely on.
Production checklist
Before launch, confirm:
- The selected model ID appears in the account used by that environment.
- The request body matches the selected API family.
- Streaming, task polling, or callbacks are tested when the route requires them.
- A fallback model or user-facing failure path is documented.
- Logs do not expose API keys, private prompts, uploaded images, or generated private assets.