Connect SillyTavern
to RawModels

What you need

• A running SillyTavern installation.
• A RawModels account, an API key and enough balance for the test request.
• An available chat model from the model catalogue. OpenJev uses a different API and is not a chat model.

1. Configure the API connection

Open API Connections in SillyTavern. Choose the Chat Completion API and the Custom (OpenAI-compatible) source. Enter these values:

API type

Chat Completion

Chat Completion Source

Custom (OpenAI-compatible)

Endpoint / base URL

https://www.rawmodels.ai/v1

API key

Your RawModels key from API key settings

Model

antislop or glm-5.3-raw, if available to your account

Include /v1 in the base URL. Do not add /chat/completions: the frontend adds that path. If the model dropdown is empty, enter the exact API ID manually. Keep your key private.

2. Choose a model and send a test

Qwen3.8 27B Antislop is the writing-focused fine-tune. GLM-5.3 Raw is agentic-first, with tool calling for multi-step workflows. Check current model prices and limits before testing.

Click Test Message in SillyTavern. Start with a short prompt and a modest response limit. Once it works, try an existing character and compare the model with your usual settings. An API format match does not make every preset or sampling option behave identically across models.

Keep the configured context and response budgets within the selected model's limits. Tool use also depends on frontend settings: SillyTavern's post-processing options marked “no tools” remove tool calls.

3. Check the cost of a real conversation

SillyTavern assembles the character, lorebook entries and conversation into each prompt. A short new message can therefore have a long billed input. Regenerating an answer makes another request. Compare the total cost of replies you keep, rather than only the price of one generated reply.

RawModels chat and API share a prepaid balance. Check usage and the current model prices as you test. A provider's context limit is not a promise that every detail in a long conversation will be recalled.

Troubleshoot a connection

401: authentication failed

Check that you copied the RawModels key correctly and that it has not been revoked. The key belongs in the API key field.

402: balance required

Check your account balance and billing. The frontend's own subscription, if any, does not fund RawModels requests.

404 or model unavailable

Check the base URL and model ID against the catalogue. Do not use a display name in place of the ID. Models marked for early access may require account approval.

429 or a slow first response

Reduce parallel requests if you hit a rate limit. Models started on demand can take longer on the first request; allow the timeout described in the API guide. Consult the model's availability details rather than assuming every request is warm.