How to connect Open WebUI to CheapTokens

Open WebUI is a self-hosted chat interface that can talk to any OpenAI-compatible API. CheapTokens exposes exactly that kind of API, so you can use it as the engine behind your chat UI and pay from one prepaid balance instead of keeping a separate subscription for each model.

By the end of this guide you will have a working connection, three models in the model picker and a first message answered.

What you need#

• A running Open WebUI where you have admin rights.
• A CheapTokens API key and a topped-up balance. If you do not have a key yet, see Your first API request with CheapTokens.
• The base URL: https://api.cheaptokens.dev/v1.

Step 1. Open the connection settings#

In Open WebUI go to Settings → Admin → Connections. In the Manage OpenAI API Connections section click Add Connection.

Menu labels are taken from the Open WebUI documentation, which does not show a version or date. If your version words them differently, look for the same section: OpenAI-compatible connections.

Step 2. Fill in the fields#

Model IDs for the allowlist:

deepseek-ai/DeepSeek-V4-Flash-0731
zai-org/GLM-5.3-Flash
MiniMaxAI/MiniMax-M2.7

Add them one by one. The allowlist keeps the picker short: only the models you list will appear for your users.

Step 3. Verify and save#

Click Verify Connection. Open WebUI checks the provider's /models endpoint, and some providers implement that endpoint differently or not at all. Saving does not test anything by itself.

So treat verification as a hint, not as the final word: if it complains but the URL and key are correct, keep the model IDs you entered manually, click Save and test with a real chat in the next step. A chat message is the reliable check.

Step 4. Send the first message#

Open a new chat, choose deepseek-ai/DeepSeek-V4-Flash-0731 in the model picker (it may be shown under this ID) and write something short, for example "Explain what a token is in two sentences." If an answer streams in, the connection works. The request is billed from your balance.

If something fails, the connection errors guide lists the usual causes: a wrong URL ending, an extra space in the key, an empty balance or a model ID with a typo.

Which model to pick in a chat UI#

All three models share one balance, so you can switch in the picker without topping up anything else:

• DeepSeek-V4-Flash (release 0731) is a good default for everyday chat. It allows up to 32,768 output tokens per answer, so long replies are less likely to be cut off, and has a 380K context window.
• GLM-5.3-Flash has a 390K context window, useful when you paste long documents. Answers are limited to 8,192 tokens.
• MiniMax-M2.7 is a third option with a 200K context and 8,192 output tokens per answer.

On output tokens, which usually make up the longer part of a chat reply, all three are priced below OpenRouter's reference price. Current prices:

What a day of chatting costs#

Chat is billed by tokens, not by message. As an example, take a day with about 200,000 input tokens and 50,000 output tokens in total. The table below is computed from the current prices and compared with OpenRouter.

The same workload on MiniMax-M2.7 costs $0.0175, and its input and output prices are $0.05 and $0.15 per 1M tokens.

Why it fits a chat UI#

• One prepaid balance for every model in the picker.
• Fixed per-model prices with no peak hours.
• OpenAI-compatible API, so Open WebUI needs no plugins.

When the balance runs low, top it up and the connection keeps working without any change in Open WebUI.

Quick checklist#

1. Settings → Admin → Connections → Add Connection.
2. URL https://api.cheaptokens.dev/v1, your API key, model IDs added manually.
3. Verify Connection, then Save.
4. Pick a model in a new chat and send a message.

Ready to try it? Get an API key, top up your balance and add the connection in a couple of minutes.

Endpoints, model ids and limits for every model are in the CheapTokens docs.