Nbility continuously updates its model catalog and routes. Instead of relying on a static “latest models” table, use these live sources:
Call GET /v1/models with an API key to retrieve the catalog currently available to that key:
curl https://api.nbility.ai/v1/models \
-H "Authorization: Bearer YOUR_API_KEY"
See the Models API for details.
Choose by workload
| Workload | Prioritize |
|---|---|
| Daily chat and code changes | Balance of quality, tools, latency, and price |
| Complex plans and long tasks | Reasoning, context window, and stability |
| Bulk summaries or classification | Unit price, throughput, and rate limits |
| Image understanding | Actual image-input support |
| Image generation | A dedicated model exposed by Images or Gemini image APIs |
| Realtime voice | Realtime model and matching event protocol |
| Embeddings / rerank | Vector dimensions, language coverage, and input limits |
A model returning text does not prove that it supports tools, images, structured output, or a client's Agent mode. Verify the capabilities you actually need.
IDs and groups
- The request
modelmust use the complete catalog ID with the same case and prefix. - A key group controls model access and billing routes. Different keys on one account may expose different catalogs.
- A client's built-in name may not equal the Nbility route ID. Use the Nbility catalog for custom providers.
- Preview IDs, dated versions, and aliases can migrate or retire. Production systems should refresh the catalog and keep a controlled fallback.
Production checklist
- Create a separate key per application and restrict its model scope.
- Test the exact capabilities you need: non-streaming, streaming, tools, and multimodal input.
- Record the response model, request ID, latency, and usage for route troubleshooting.
- Keep real model IDs in configuration; do not treat “latest” or “flagship” labels as stable contracts.