SillyTavern

SillyTavern connects to any OpenAI-compatible Chat Completions API through its custom endpoint source. Our abliterated models don't refuse adult roleplay and creative writing. Models labelled Official DeepSeek API are DeepSeek's stock models and keep DeepSeek's own refusals, so pick an abliterated model for adult roleplay. See Models.

Placeholders: $API_BASE is https://api.abliteratedllm.dev/v1, $API_ORIGIN is https://api.abliteratedllm.dev, and $YOUR_KEY is your API key.

Connect

1. Open API Connections (the plug icon in the top bar).
2. Set API to Chat Completion.
3. Set Chat Completion Source to Custom (OpenAI-compatible).
4. Enter the custom endpoint URL: $API_BASE. It must end in /v1; don't add /chat/completions.
5. Enter your API key: $YOUR_KEY.
6. Click Connect. SillyTavern reads our /v1/models endpoint, so you can pick glm-5.3-flash-abliterated, deepseek-v4.1-flash or deepseek-v4-pro from the model list.
7. Click Test Message to confirm it works.

Recommended settings

• Streaming works and shows tokens as they arrive.
• Max response length can go up to the model's max_output_tokens from GET /v1/models, 65536 for our abliterated models. Trial accounts are capped at 4096 per response.
• Context size can be set as high as the model's context_length from GET /v1/models, for example 262144 tokens for glm-5.3-flash-abliterated. Longer context costs more per message, because the whole chat history is sent each time. Prompt caching cuts the cost of the repeated part.
• Prompt Post-Processing can stay at None. If a character card or preset produces errors about message order, try Semi-strict.

Good to know

• Text only. Images attached to messages fail with 400 unsupported_content.
• 18+ only. See our Acceptable use policy.

Verified against SillyTavern docs on 2026-11-03: Chat Completions, custom OpenAI-compatible endpoint.