FAQ
Answers common usage questions and explains protocols and default models.
Q1: What is this service?
A: This is an API gateway that aggregates AI models. It natively supports Claude Code channels over the Anthropic protocol and can connect to OpenClaw, VS Code extensions, Claude Code, OpenCode, Cherry Studio, and other platforms.
Q2: Does it use the OpenAI protocol?
A: This channel uses the Anthropic protocol, so select Claude-compatible / Anthropic-compatible when connecting it. It is also compatible with the OpenAI protocol.
Q3: Why is the default model listed as claude-sonnet-4-6?
A: This ensures compatibility with the way Claude Code channels make requests. After connecting successfully, you can switch to any supported model name.
Q4: Which models are supported?
A: The service supports many model families, including OpenAI, Claude, Zhipu, Kimi, Qwen, Tencent Hunyuan, MiniMax, StepFun, Meituan LongCat, Doubao, DeepSeek, and Xiaomi MiMo. The newly added claude-sonnet-5, claude-fable-5, Tencent Hunyuan, and LongCat-2.0 models have also been added to the public support list. For current model names, multipliers, and capabilities, see the model list and model pricing and billing guide.
Q5: How is pricing calculated?
A: Billing is based on each model's multiplier. The base rate is 1x = RMB 40 per 100 million tokens = RMB 0.4 per 1 million tokens. Rates for other models scale linearly with their multipliers, and cache hits usually cost less. See the model pricing and billing guide for details.
Q6: How are plans calculated?
A: Your available credits are based on the plan shown on the order or top-up page. Credits never expire or reset, so unused credits remain available. If you already have a Key, go directly to the top-up page to add more credits. Contact support only if you need a custom plan.
Q7: Is caching supported?
A: Yes. Cache hits reduce overall usage costs. Cache hits are billed at 0.1x.
Q8: What use cases are supported?
A: Supported use cases include coding CLIs, VS Code, IDE extensions, agents, customer support, web scraping, quantitative analysis, education, workflows, and e-commerce AI.
Q9: Can I use the service both inside and outside mainland China?
A: Yes. There are no domestic or international network restrictions.
Q10: Are custom plans available for high-volume usage?
A: Yes. High-volume users can request custom pricing. Contact support for more information.
Q11: Why does Codex keep showing Reconnecting?
Try another node. In most cases, Reconnecting... appears because Codex initially uses a ws request. Without tun mode enabled on your proxy, ws often cannot connect, so Codex switches to https after 5 failures. If all 5 attempts fail, first use CC-Switch to check whether your endpoint responds to ping.
Q12: Error "unknownAPI Error: Content block not found"
If you still see ❌ Error: unknownAPI Error: Content block not found after restarting while a proxy is enabled, disable the proxy. The request should work after the proxy is turned off.