Frequently Asked Questions

Visit the platform website, click the "Register" button, and fill in your username and password to register. If the platform has email verification enabled, you will also need to provide your email address and verification code. After successful registration, the system automatically creates an initial API Key.

Click "Forgot Password" on the login page, enter your registered email to receive a verification code, then set a new password. If you cannot reset via email, please contact the administrator.

After logging in, go to "2FA Settings" in your account center. Use Google Authenticator or another TOTP app to scan the QR code, then enter the verification code to complete binding. Once 2FA is enabled, you will need to enter a dynamic verification code each time you log in. Please safely store the backup recovery codes.

You can choose to delete your account in the account center. Note that after deletion, all API Keys will become invalid and your remaining balance will not be refunded. Please confirm before proceeding.

The platform supports Creem online payment and redemption code top-up.

The minimum top-up amount is determined by the platform settings, typically 1 USD or equivalent in RMB. Please refer to the minimum amount displayed on the top-up page.

Under normal circumstances, payment arrives within 1-2 minutes after completion. If it takes significantly longer, please check whether the payment was successful, or contact the administrator with proof of payment for manual processing.

Log in to the console and go to "Usage Logs" to view detailed records of each API call, including call time, model used, Token consumption, cost, etc. You can also view consumption trend charts in the "Dashboard".

Generally, account balances do not expire. However, promotional credits granted by the platform may have an expiration date. Please follow platform announcements.

After logging in, go to the wallet page, find the redemption code input field, paste the redemption code, and click Redeem. Upon successful redemption, credits are automatically added to your account. Each redemption code can typically only be used once and has an expiration period.

Token consumption = Input Tokens + Output Tokens. Input Tokens are the content you send to the model (including system prompts, conversation history, and user messages). Output Tokens are the model's generated response. Different models have different per-token prices; check the pricing page for details.

The API address is the base URL provided by the platform, for example https://api.shallowseek.top/. For the specific address, please check the platform site list or consult the administrator.

Three steps to get started: (1) Register or log in to the platform and get your API Key (starts with sk-); (2) Set the API Base URL to the platform-provided address; (3) Use the OpenAI SDK or send HTTP requests directly. For detailed code examples, refer to the "Using AI Models" page.

Yes. Set stream to true in the request parameters to enable SSE (Server-Sent Events) streaming output. With the OpenAI SDK, simply set stream=True.

Yes. For models that support Function Calling (such as gpt-4o, claude-3.5-sonnet, etc.), you can pass tools and tool_choice parameters in the standard OpenAI format. The model will return a function call response.

401 indicates authentication failure. Possible causes: (1) The API Key is incorrect or does not exist; (2) The Key has expired; (3) The Key has been disabled; (4) The Authorization header format is incorrect (correct format: Bearer sk-xxx).

429 indicates that the request has been rate-limited. Possible causes: (1) The request frequency exceeds the rate limit; (2) The Key's quota has been exhausted. Reduce the request frequency or top up credits before retrying.

The -c suffix means "per-request billing." For example, gemini-2.5-pro and gemini-2.5-pro-c call the exact same model; the difference is that the former charges by Token, while the latter charges a fixed fee per call. Choose the appropriate billing method based on your use case.

The platform implements global rate limiting. If requests are too frequent, a 429 error will be returned. We recommend implementing an exponential backoff retry strategy on the client side. The specific limit values are subject to platform settings.

(1) Check network connectivity (ping/traceroute to the API host); (2) Try switching API endpoints (CDN endpoint / primary endpoint / backup endpoint); (3) Check firewall and proxy settings; (4) Increase the client timeout setting.

No. All API endpoints are designed for direct connection. Using a proxy or VPN may actually cause unstable connections or be flagged as abnormal traffic. Please connect directly.

Technically yes, but exposing your API Key in frontend code is strongly discouraged. If you must use it in a browser, we recommend forwarding requests through a backend service, or using the platform's Playground feature for testing.

Ensure your client has the correct Content-Type and Accept-Charset settings. Use UTF-8 encoding. If using Python, make sure to set response.encoding = 'utf-8'.

Most third-party clients that support the OpenAI API format can be used directly. Simply change the API Base URL in the client settings to the platform-provided address and enter your sk- Key as the API Key. Common clients include ChatBox, Cherry Studio, LobeChat, NextChat, and more.

The platform aggregates 40+ leading model providers, including OpenAI (gpt-4o, gpt-4.1, etc.), Claude (sonnet, opus, etc.), Gemini (pro, flash, etc.), DeepSeek, Qwen, as well as various specialized models (image generation, embeddings, reranking, etc.). Check the console pricing page or the "Using AI Models" page for details.

If a model is not visible in the model list, it usually means the model is currently unavailable (upstream account shortage or temporarily suspended). You can try using other similar models as alternatives. If you urgently need a specific model, contact the administrator to check its status.

Top-tier models (gpt-4o, claude-opus, gemini-pro) are suitable for complex reasoning, code generation, and professional writing. Mid-range models (gpt-4.1-mini, claude-haiku, gemini-flash) are suitable for daily conversations, simple Q&A, and translation. Smaller models (gpt-4o-mini, etc.) are fast and cost-effective, suitable for high-concurrency simple tasks. We recommend choosing based on task complexity — use small models for simple tasks to save costs, and large models for complex tasks to ensure quality.

Model prices may change based on upstream provider adjustments. The platform strives to maintain price stability and will provide advance notice of any adjustments. Please follow the pricing page for the latest price information.

Multimodal models (such as gpt-4o, claude-3.5-sonnet, gemini-pro) support image input. You can pass base64-encoded images or image URLs in the messages using the standard OpenAI Vision API format.