Open WebUI is a self-hosted chat interface. It connects to any provider that implements the OpenAI chat-completions protocol, and BroadNet Inference does. The steps follow Open WebUI’s guide for OpenAI-compatible providers.
Setup
-
Get a BroadNet key. For a chat UI that several people use, a pay-as-you-go key is the right fit. See API keys and IP allow-lists.
-
Open the connections page. Go to Settings > Connections in the admin section and find Manage OpenAI API Connections.
-
Click Add Connection and fill in:
Field Value URL https://inference.broadnet.ai/v1(no trailing slash)API Key your BroadNet key Model IDs add each ID below, one at a time, with the + icon -
Add the model IDs by hand. Type
broadnet/glm-5.3, click +, then repeat for the others you want. BroadNet’s docs describeGET /v1/modelsas public catalog metadata for monitors, so don’t rely on it to fill the model picker. -
Click Save. Open WebUI’s docs note that saving doesn’t test the connection.
-
Start a chat and pick the model from the model selector.
With environment variables
If you run Open WebUI in Docker, you can set the connection at startup:
OPENAI_API_BASE_URL=https://inference.broadnet.ai/v1
OPENAI_API_KEY=your-broadnet-key
To add BroadNet next to an existing OpenAI connection, use the plural forms, separated by semicolons: OPENAI_API_BASE_URLS and OPENAI_API_KEYS. OPENAI_API_CONFIGS sets per-connection options such as model_ids, keyed by the connection’s index in that list:
OPENAI_API_CONFIGS='{"0": {"enable": true, "model_ids": ["broadnet/glm-5.3", "broadnet/qwen3.8-27b"]}}'
Open WebUI’s environment variable reference marks these as ConfigVar settings. They are read on the first launch and saved in the database. After that, the admin panel wins. Edit the connection in Settings > Connections in the admin section, or set ENABLE_PERSISTENT_CONFIG=False if you want the environment to apply on every start.
Which model to pick
| Use | Model | Context |
|---|---|---|
| General chat, tools, JSON | broadnet/glm-5.3 |
1,048,576 |
| Image questions, Arabic | broadnet/qwen3.8-27b |
262,144 |
| Long documents, images | broadnet/deepseek-v4.1-flash |
1,048,576 |
| Low cost, free credit | broadnet/gemma-4-26b-a4b |
32,768 |
BroadNet tested Qwen3.8 27B in Arabic, including Gulf and Egyptian dialects; see choosing an LLM for Arabic. Gemma 4 VL 26B is the only chat model that works on the $5 starting credit; the credit also covers the two Transcribe models. Prices for each are on the pricing page.
Open WebUI passes standard parameters such as temperature and max_tokens, and supports tool use when the model and server support tools. BroadNet lists tools for GLM 5.3 and Qwen3.8 27B.
Troubleshooting
- Verify Connection fails with 400, 401 or 403. Open WebUI’s verify button calls
/models. Its docs say chat completions can still work. Add the IDs to the allowlist and test a chat. If chat also returns 401, the key is wrong, expired, or out of pay-as-you-go allowance. - The model is listed but the first message fails. Check that the ID matches BroadNet’s exactly, with the
broadnet/prefix. GLM 5.3, Qwen3.8 27B and DeepSeek V4.1 Flash need a top-up. - Empty replies. Reasoning models (DeepSeek V4.1 Flash, GLM 5.3, Qwen3.8 27B) spend output tokens on reasoning first. If you set max_tokens in the model’s Advanced Params, raise it or clear it. Why answers come back empty.
- 429 when several people chat. A pay-as-you-go key allows 30 requests a minute and 4 at once. Open WebUI also makes background calls, such as chat titles, on the same key. See errors and retries.
- Changed env vars do nothing. The first launch saved the old values. Edit the connection in the admin panel, or see the environment variable section above.
- Using a plan key. An unlimited plan session runs one request at a time, and keys can’t be shared. It suits one person’s own instance; extra users’ requests queue.
