← Help CenterDescription
Pricing, protocol, filename, throttling, error reporting - it's all here.
Personal ¥30/30 days(Alipay), or$5/30 days（USDT / Stripe）。 Small team from ¥300/30 days, including multiple accounts and invoicing. For details, seePrice html。
No. When it expires, it stops, you don't need to notify us, and no money will be deducted from your account.
The ¥30 bracket includes$4 usage. It won’t be cut off immediately after use – you can refill and continue, or wait until next month.
Because it can't be done. If you write "unlimited", you will either use it to secretly limit the speed, or the event will be canceled at any time. We would rather put the limit on the bright side: you know how much you have before buying, and we also know how much we have to bear.
If you are using a command line client, just runnpx newcclm --status;
Or go toQuery UsageSee the page.
If you use a small amount within 24 hours after activation, you can contact us for a full refund. If you have used it for a period of time, there will be no refund - because the credit has already incurred actual costs.
Yes. Just send us the invoicing information (name, tax number, amount). Both individual and small team files can be opened.
OpenAI compatible. Do not choose Anthropic native - it will not connect directly.
https://sccclm.com/v1A few tools require the complete endpoint to be filled in. In that case, fill in thehttps://sccclm.com/v1/chat/completions。
Do not overlap/v1(Will become/v1/v1）。
. Pass"stream": true needed.
1M token. The actual available length is limited by the client itself, subject to the client display.
API 层面有四个模型名，按事情挑：
| Model Name | Purpose | When to use |
|---|---|---|
newcclm |
Daily Main Power | 不确定选哪个就用它。回邮件、整理资料、一般提问 |
newcclm-fast |
更快 | 要速度、不需要想太久的事 |
newcclm-coding |
Encoding and long context | 写代码、读长文档、跑 Agent |
newcclm-free |
免费 | 试用量、零成本任务 |
不用。客户端把它收敛成了两个模式：对话里打 /fast or /smart 就行，
你不用记模型名。四个模型名是给「填配置的场合」用的 ——
比如 Cherry Studio、Cursor、DSH 这些工具的模型栏。
因为大部分人不需要在十几个型号之间做选择题。我们把它收敛成两种：
&Fast(return email, change copy, ask questions) andTingkatkan（分析、读长文档、写代码）。
想更细就用 newcclm（日常主力）和 newcclm-free（免费）。
Not in the command line client - you just need to select the gear. If you have specific needs (such as connecting to internal systems), Contact us to talk individually.
gpt-4o or claude-*Such a name?
No. We only recognize our own stall name:newcclm、newcclm-fast、newcclm-coding、newcclm-free.
If you fill in the real model name, it will report that the model does not exist.
我们把底层的型号公开写在 Model table on the homepagemile
| 底层模型 | Purpose | 对应我们的档位 |
|---|---|---|
| GLM-5.3-Flash | Daily Main Power | newcclm / newcclm-fast |
| Qwen3.8-Flash | Encoding and long context | newcclm-coding |
| 免费模型 | 零成本模型 | newcclm-free |
| DeepSeek V4-Flash | 备用线路，主线路故障时接管 | 自动切换，不用你选 |
都是 1M 上下文。
No. If you choose "Enhanced", go to Qwen, if you choose "Quick" or "Daily", go to GLM, Does not downgrade to another gear。
DeepSeek only takes over if the main line fails - this is disaster recovery, not degradation. Your wait time and result quality won't drop when you switch.
要。 Each model has its own terms of use, you need to abide by them when using our services ——
This includes not being used to generate illegal content, not being able to resell the model itself, etc.
If you are not sure whether a certain use is allowed or not, ask us first, don’t try it directly.
Yes. In order to fairly allocate resources, personal files have a request frequency limit for a single account. It cannot be touched during normal use; if you want to run batch tasks, tell us in advance and we will adjust it for you.
We have prepared more than one upstream line. If there is a problem with the main line, it will be automatically switched to avoid "suddenly not being able to use it".
Let’s be honest: no one in this industry can guarantee 100% non-stop service. What we can promise isWhen problems arise, we will deal with them as soon as possible.
and tell you the truthinstead of letting you guess.
Possibly. Slow down and try again later; if it lasts long, let us know.
| What you see | What is the problem | How should I do? ! |
|---|---|---|
Invalid token |
The key is incorrect, missing, or invalid | Check if the key is replicated completely; command-line client logs back in |
model not found |
Model name misspelled | Can only be filled innewcclm / newcclm-fast / newcclm-coding / newcclm-free |
404 |
Wrong address | Confirm Yeshttps://sccclm.com/v1and does not overlap/v1 |
Mentionsquota / Balance |
Dosage ran out | Contact us to add or wait for next month |
429 |
Too many requests, throttling triggered | Decrease concurrency or try again later |
401andupstream |
Our upstream problem, not yours | Try again later; keep telling us this |
500 / 502 / 503 / 504 |
我们这边或上游临时出问题 | 等几秒重试；连续超过几分钟就告诉我们 |
先别改配置。500 到 504 这类错误
几乎都不是你填错了 —— 是服务端或上游临时出了问题。
你改配置反而会把本来能用的设置弄乱。
等几秒重试一次。大部分 5xx 是瞬时的，重试就好了。 你的客户端一般会自动重试几次，不用手动干预。
连续几分钟都是 5xx —— 那就是真出问题了，告诉我们一声。 我们准备了不止一条上游线路，主线路故障会自动切到备用线路， 但这个切换不是瞬间完成的。
5xx → 服务端的问题，不是你的配置，等一会儿再试401 / 403 → 密钥或权限的问题，要检查你的 Key
both_paths_failed这通常不是故障，是你把提供方加错了位置。
有些客户端（DSH 这类）加了提供方之后会同时探测两条路径：
OpenAI 的 /v1 和 Anthropic 的 /anthropic/v1。
我们只提供 OpenAI 那条 —— 所以 Anthropic 那条会 404，
两边一综合，客户端就报 both_paths_failed。
怎么办：把协议锁成 OpenAI 兼容（openai-completions），
并且确认你没有把它加进 DeepSeek 或 Anthropic 的官方目录里
—— 那些目录自带 Anthropic 协议，填我们的地址也连不上。
具体见DSH 配置指南的「概述」一节。
upstream 字样。
有 upstream 就是我们这边的事，你不用折腾自己的配置；
没有的话，八成是协议或地址填错了。
in the dialogue. Will not be used for training and will not be resold. We only retain the information needed for billing (how much was used and when). If you have more stringent requirements, please contact us to discuss individually.
The command line client saves the login credentials on your own computer~/.newcclm/config.json(Permissions only readable by you).
What is saved on our server is account information, and plain text passwords are not saved.
Send us these three things, and you can locate them basically once:
1. Complete prompt on the screen (screenshot is most convenient)
2. The name and version of the tool you use
3. What did you do just now?