Documentation
FAQ: Which model should I choose first? in the NexoRouter documentation.
FAQ: Which model should I choose first?
Start with a low-cost, fast chat model while you are still validating base URL, API key, model ID, and Usage Logs. Move to a larger or more expensive model only after the integration path works.
Starter choices
| Need | Start with | Why |
|---|---|---|
| API setup check | deepseek-v4-flash | Fast, low-cost first request and agent-loop testing. |
| General app features | gpt-4o-mini | Balanced text generation, extraction, and support workflows. |
| Multilingual output | Qwen/Qwen-Plus | Good first option for English, Spanish, and Chinese prompts. |
Always copy the current model ID from Models. Do not rely on old screenshots, vendor docs, or memory.
Selection checklist
- Pick the cheapest model that can complete the task.
- Confirm the model appears in
GET /v1/models. - Confirm the API key's model scope includes that model.
- Run one small non-streaming request.
- Check Usage Logs for cost, latency, status, and request ID.
- Only then test larger prompts, agent tools, or production traffic.
When to switch models
| Symptom | Try |
|---|---|
| Text quality is too weak | A stronger chat model from Models. |
| Latency is too high | A faster model or shorter prompt. |
| Cost is too high | A lower-cost model and smaller max_tokens. |
| Agent edits fail | A model with better tool-use behavior, then retest in a disposable branch. |
model_not_found | Copy the exact model ID and check key model scope. |