How to connect Open WebUI to Unquant
Open WebUI gives you a ChatGPT-style interface you host yourself. Add Unquant as a connection and you get uncensored frontier models behind it, with no per-token bill.
1 Add a connection
In Open WebUI open Admin Panel → Settings → Connections and add one under OpenAI API:
Save. Open WebUI reads
/v1/models, so your plan's model appears in the model picker automatically.2 Set the context length
In the model's advanced parameters, set the context length to your plan's window (up to 262,144 tokens) so long conversations and documents are not cut short.
Images and documents
- Visionboth models accept images. Open WebUI sends uploads inline as base64 (PNG, JPEG or WebP), which is what Unquant accepts. Image links by URL are not fetched.
- Documentswith a 262K window you can often paste a whole document instead of relying on retrieval snippets.
Sharing with a team
Everyone on your Open WebUI instance shares your plan. Your plan's concurrent streams are how many answers generate at once; extra requests get a 429 rather than slowing everyone down, so pick a tier with enough streams for the people using it at the same time. See plans.
Troubleshooting
The URL must end in /v1, and the key must belong to an active plan.
An image was sent as a link; upload the file instead.
Paste the full key again; revoked or mistyped keys are rejected.
The model ID doesn't match your plan: GLM plans use glm-5.3-flash, Qwen plans qwen-3.8-flash-next.
Keep reading
- Call Unquant from Python and JavaScript with the OpenAI SDKUse the official OpenAI Python and JavaScript SDKs with Unquant: base URL, streaming, images, tool calling and reasoning control, with copy-paste examples.
- Why the context window matters (and when 262K is worth it)What a 262K-token context window actually buys you in roleplay, coding, and agents, how providers price it, and when the full window is worth it.
- Uncensored AI for coding and agentsWhy developers use an uncensored model for coding and agents: no refusals on security work, reverse engineering or edgy content pipelines, plus 262K context and unlimited tokens for long agent loops.