❓ Frequently Asked Questions

TOENK API is an AI model aggregation platform that provides unified access to 20+ leading AI models through a single OpenAI-compatible API endpoint. Simply change your base_url to https://toenk-api.com/v1 and instantly access models from DeepSeek, OpenAI, Anthropic, Google, Alibaba (Qwen), Zhipu (GLM), Moonshot (Kimi), ByteDance (Doubao), and more.

Key advantages include optimized mainland China routing with <100ms latency, competitive pricing starting at ¥0.5/1M tokens, and full compatibility with existing OpenAI SDK code.

Yes, fully compatible. TOENK API implements the OpenAI Chat Completions, Embeddings, and other major interface protocols. You only need to change base_url to https://toenk-api.com/v1 — your existing code works without any modification. We also support Anthropic SDK and Google AI SDK interface conventions for their respective models.

Extremely fast. Our servers are hosted on BGP-optimized connections with direct peering to China Telecom, China Unicom, and China Mobile:

• P50 latency: <100ms
• P99 latency: <500ms
• SLA: 99.9% uptime guaranteed
• Multi-node cluster: automatic failover with zero-downtime transitions

We offer multiple support channels:

• 📧 Email: support@toenk-api.com — general inquiries, technical issues, billing
• 💬 Ticket System: Submit tickets directly from your dashboard after login
• 📖 Documentation: Visit toenk-api.com/en/docs/

Our support team operates 7×24. Urgent issues receive an initial response within 5 minutes. Standard inquiries are answered within 2 hours during business hours.

Registration is simple:

• Go to the Registration Page
• Enter your email address and create a password (minimum 8 characters with mixed case)
• Click the verification link sent to your email to activate your account
• Log in and start using the API immediately. New users receive 500,000 free tokens

On the login page, click the "Forgot Password" link. Enter your registered email address, and we'll send a password reset link. If you don't see the email, check your spam folder. If it still doesn't arrive, contact support for manual reset assistance.

After logging in, navigate to the Token Management page at /keys in your dashboard. Click "Add Token", give it a name, set usage limits, and generate. Copy and save the key immediately — it will not be displayed again after you navigate away from the page.

Absolutely. You can create separate API Keys for different environments — for example, one for development, one for production, and one for testing. Each key supports independent quota limits and IP whitelist settings, giving you granular access control.

Log in and go to the Profile Settings page at /profile. You can update your display name, email preferences, and notification settings. Sensitive changes (such as changing your email address) may require email verification for security purposes.

You are billed per token based on actual usage — no monthly or annual subscription fees. Prices vary by model. For a complete breakdown, check our Model Pricing Table.

Billing is denominated in RMB (CNY). As a reference, DeepSeek V4-Flash costs just ¥0.5 per 1M tokens. New users receive 500,000 free tokens upon registration — check your remaining balance in the dashboard.

Yes! Every new user receives a 500,000 token free quota upon registration. To put that in perspective, at DeepSeek V4-Flash pricing (¥0.5/1M tokens), this equates to roughly 1 million short conversations. The free quota is valid for 30 days and any unused portion expires after the validity period.

After logging in, go to the Top-Up page in your dashboard. We support the following payment methods:

• 🧧 Alipay: Scan the QR code for instant credit
• 💚 WeChat Pay: Scan the QR code for instant credit

Topped-up balances never expire. For bulk purchases (monthly spend over ¥1,000/CNY), contact support to arrange VIP rates and a dedicated quota pool.

Log in and go to the Usage History page in your dashboard. You can view detailed token usage statistics and billing records. Filters are available to narrow down results by time range and model type.

When both your prepaid balance and free quota are depleted, API requests will return 402 Payment Required or 403 Insufficient Quota errors. Service resumes immediately after a top-up. We recommend setting a balance alert threshold in your dashboard — the system will automatically notify you when your balance drops below the configured level.

Getting started takes just three steps:

• 1. Register at toenk-api.com and verify your email
• 2. Generate an API Key from the Token Management page in your dashboard
• 3. Switch your base_url to https://toenk-api.com/v1 in your existing OpenAI SDK setup

That's it — your existing code works with zero modifications. All new accounts come with 500K free tokens to try everything out.

base_url (also called api_base in some SDKs) is the parameter in OpenAI SDKs that specifies the API server address. When using TOENK API, simply set it to:
base_url = "https://toenk-api.com/v1"

Python example:

from openai import OpenAI

client = OpenAI(
    api_key="sk-your-key",
    base_url="https://toenk-api.com/v1"
)

response = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Hello!"}]
)

Simply change the model parameter in your API request. Some examples:

• "model": "deepseek-v4-flash" → DeepSeek V4-Flash
• "model": "gpt-4o" → GPT-4o
• "model": "claude-sonnet-4" → Claude Sonnet 4
• "model": "gemini-2.5-flash" → Gemini 2.5 Flash
• "model": "qwen3.6-plus" → Qwen 3.6 Plus

Get the full model list via the GET /v1/models endpoint (requires API Key).

Standard rate limits are 60 requests per minute per model. This is a per-model limit, so using multiple models in parallel effectively increases your throughput. If you need higher limits for production workloads, contact support to discuss dedicated capacity options.

Yes. We fully support OpenAI-compatible SSE (Server-Sent Events) streaming. Set stream: true (or stream=True in Python) in your request to enable it. First-token latency is extremely low, making streaming ideal for real-time chat and interactive applications.

Absolutely. TOENK API fully supports OpenAI-compatible Function Calling and Tool Use interface specifications. Define your tools, and the model will automatically decide when to invoke them. This works perfectly with Agent frameworks, code execution tools, and external API integrations.

Yes. Vision-capable models include gpt-4o, gpt-4o-mini, claude-sonnet-4, gemini-2.5-flash/pro, and others. Use the same format as the OpenAI Vision API — pass images as base64-encoded URLs in the messages.content array. Audio input is supported on select models such as gpt-4o-audio-preview. File uploads for document analysis are also supported on applicable models.

Our platform supports 26+ models across 6 leading AI providers. Here's the lineup:

• DeepSeek: deepseek-v4-flash, deepseek-v4-pro, deepseek-reasoner, deepseek-chat, deepseek-v3.2
• OpenAI: gpt-4o, gpt-4o-mini, gpt-5, gpt-5.4
• Anthropic: claude-sonnet-4, claude-opus-4
• Google: gemini-2.5-flash, gemini-2.5-pro
• Qwen (Alibaba): qwen3.6-plus, qwen3.6-flash
• GLM (Zhipu): glm-5, glm-5.1
• Other: grok-4.1 (xAI), kimi-k2.6 (Moonshot), doubao-seed-2.0-pro (ByteDance), ernie-5.0 (Baidu), and more

For a complete list with pricing, see our Model Pricing Table or call GET /v1/models.

Here's a quick guide by use case:

• Daily chat / Light tasks: deepseek-v4-flash (best value, ¥0.5/1M tokens)
• Complex reasoning / Math / Code: deepseek-v4-pro or deepseek-reasoner
• Long context / Big documents: gemini series (1M context window) or deepseek series
• Creative writing: claude-sonnet-4
• Cost-effective general use: gpt-4o-mini or qwen3.6-flash
• Chinese language optimization: qwen series, glm series, doubao series

All major models support Function Calling for Agent/ tool-use workflows.

Yes, we constantly monitor the AI landscape and add new models as they are released. We prioritize high-performance, cost-effective models from both established and emerging AI labs. Model additions are announced via release notes and in-dashboard notifications. For the latest list, call GET /v1/models at any time.

Context lengths vary by model family:

• DeepSeek series: 128K tokens
• GPT-4o / 4o-mini: 128K tokens
• GPT-5 / 5.4: 256K tokens
• Claude series: 200K tokens
• Gemini series: 1M tokens (largest available)
• Qwen series: 128K tokens
• GLM series: 128K tokens
• Other models: typically 128K–200K tokens

Our competitive pricing comes from three advantages:

• Volume aggregation: By pooling demand across thousands of users, we negotiate lower wholesale rates with upstream providers
• Optimized routing: Smart load balancing and model caching reduce per-request overhead
• Lean operations: Automated infrastructure keeps our operating costs minimal

We pass these savings directly to you. Our transparent pricing page shows exact per-model rates with no hidden fees.

Follow these best practices to keep your API Key secure:

• Never store in plaintext: Use environment variables or a secrets manager (e.g., Vault, AWS Secrets Manager)
• Use IP whitelists: Bind API Keys to specific IP addresses so only requests from trusted sources work
• Set quota limits: Configure daily and total usage caps per Key to prevent unexpected overage
• Rotate regularly: Generate new Keys every 90 days and retire old ones
• Never share publicly: Don't expose Keys in client-side code, public repositories, or web pages

Yes. All API communications are encrypted using TLS 1.2 / TLS 1.3 with certificates issued by Let's Encrypt. Authentication uses Bearer Token headers — every request must carry a valid API Key. HSTS is enforced across all endpoints, ensuring HTTPS-only connections at all times.

We strictly follow data minimization principles:

• Request content: We log metadata only (timestamp, model used, token count). Conversation content is not stored on our servers
• Access logs: Retained for 7 days for operational troubleshooting, then automatically purged
• API Keys: Stored as one-way hashes — even if data were compromised, keys could not be reconstructed
• Data sharing: We never sell or share user data with third parties

A Data Protection Agreement (DPA) is available on request for enterprise customers requiring additional assurances.

Yes. We operate a multi-layered security framework:

• WAF rules: Automatically blocks SQL injection, XSS, path traversal, and other OWASP Top 10 threats
• Rate limiting: Per-IP request caps prevent abuse
• IP blacklisting: Automated detection and blocking of malicious scanners and attack sources
• Channel redundancy: Multi-node cluster with automatic failover — a single node failure causes zero service disruption

Unused prepaid balances are refundable. If you have topped up funds that you haven't used, contact our support team at support@toenk-api.com with your account details and refund request. Refunds are processed within 5-7 business days. Note that promotional/free credits and already-consumed usage are not eligible for refund.