# Passion8 FAQ: API Keys, Models, Connections and Billing

> Common Passion8 questions about Base URLs, models, billing, authentication and differences from direct provider access.

URL: https://docs.passion8.cc/en/docs/faq
Language: en
Publisher: Passion8
Last updated: 2026-10-11

Find common connection questions here. For anything else, see your client's troubleshooting page or contact support. Invitation links and sharing text are in [referral rewards](https://docs.passion8.cc/en/docs/invite).


### 


### Should the Base URL include /v1?

It depends on the client:

- **Gemini CLI**: `https://passion8.cc`, **without** `/v1`.
- **Claude Code**: `https://passion8.cc`, **without** `/v1`.
- **Codex / Grok / OpenAI-compatible clients**: `https://passion8.cc/v1`, **with** `/v1`.

An incorrect `/v1` suffix is a common cause of failed connections.




### Which models are supported?

Use the models actually available in the [console catalog](https://passion8.cc), covering Anthropic (Claude), OpenAI (GPT / Codex), Gemini and other providers. Copy the exact model ID from the catalog.




### Is billing the same as the official provider?

Charges are converted proportionally against official pricing with no hidden fees. Usage, rates and caching can be reconciled in the console. See [top-ups and billing](https://docs.passion8.cc/en/docs/billing).




### How do I fix 401 or authentication failures?

Check in order:

1. The complete key was copied without extra spaces.
2. The Base URL has the correct `/v1` suffix.
3. The key is in the correct location: Claude environment/settings.json, Codex `~/.codex/auth.json`, Grok Build `~/.grok/config.toml`, Gemini CLI `~/.gemini/.env`.
4. **Restart the tool or open a new terminal** after changing configuration.




### Why does it say model not found?

Use an ID actually available in the [model catalog](https://passion8.cc). For Codex, also verify Responses API support.

If the ID is correct, check the **token group**: choose a specific group containing that model instead of assuming the default group includes it.




### Are streaming, tool calls and caching supported?

Yes. Passion8 forwards provider protocols, including streaming, tool calls and context caching, without requiring SDK changes.




### How does this differ from direct provider access?

The integration is similar: change the Base URL and key. Passion8 adds direct access from mainland China, redemption-code top-ups, multiple routes and observable request logs and usage.






## Diagnose by error

| Symptom | Check first | Next step |
| --- | --- | --- |
| 401 / invalid API key | Correct Passion8 key; stale environment overrides | Update credentials locally; never send support a complete key |
| 403 / permission denied | Token group, model access and client requirements | Check permissions and share timestamp plus a redacted error |
| 404 / model not found | Correct `/v1` suffix and model ID | Compare the [quickstart](https://docs.passion8.cc/en/docs/quickstart) and console catalog |
| 429 / rate limited | Concurrency, channel limits or balance | Reduce concurrency and honor retry hints; avoid retry loops |
| 5xx / timeout | [Service status](https://status.passion8.cc/) and local network | Record time, model and a redacted error without publishing content or credentials |

The same status code can come from different layers. This is a diagnostic order, not a definitive explanation for every failure. See [billing](https://docs.passion8.cc/en/docs/billing) for charges.