DeepSeek API in SillyTavern: the endpoint, every DeepSeek id, and the cost per message
Connect SillyTavern to DeepSeek through one Custom (OpenAI-compatible) source, pick from 15 DeepSeek ids with their Context Size, and see what one 3K-in / 300-out message costs on each at live rates.
SillyTavern reaches DeepSeek through its Chat Completion API with the Custom (OpenAI-compatible) source: a base URL, a key and a model id. On RelaySwitch the base URL is https://api.relayswitch.io/v1, the key is a sk-relay- key from the console, and the model id is any DeepSeek id in the table below: the stock builds and the builds the catalog tags uncensored or abliterated, on the same key.
Keys are prepaid from $10 by card or crypto, with no subscription and no KYC. Every price on this page is read from the live catalog when the page is built, never typed. The SillyTavern settings guide covers the samplers, Context Size and reasoning settings in full; this page covers the DeepSeek part.
01The connection panel
Open the API Connections panel (the plug icon) and set:
API: Chat CompletionChat Completion Source: Custom (OpenAI-compatible)Custom Endpoint (Base URL): https://api.relayswitch.io/v1Custom API Key: sk-relay-...Enter a Model ID: deepseek-v4.1-flash-1017-uncensored- Custom Endpoint (Base URL):
https://api.relayswitch.io/v1, ending in/v1. SillyTavern adds/chat/completionsitself. - Custom API Key: the key, with no space before or after it.
- Enter a Model ID: a DeepSeek id from the table, pasted exactly. After Connect, the Available Models dropdown fills from the key and lists the same ids.
Click Connect. A green light and the model name beside it means the key answered; a red one with 401 is the key, 404 is the id or a URL that does not end in /v1.
02Which DeepSeek id to use
Every DeepSeek id on the key, with the kind of build, the context window to set as Context Size, and the live price of one 3K-in / 300-out message and 100 of them. The builds tagged uncensored or abliterated come first: DeepSeek models with the refusal behaviour tuned out or removed from the weights, so a card plays as written without a jailbreak block in the system prompt. The uncensored models page shows each one's daily check.
| Model | Model ID | Build | Context Size | Input | Output | one 3K-in / 300-out message | 100 messages |
|---|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash 1017 Uncensored | deepseek-v4.1-flash-1017-uncensored | Uncensored | 1M tokens | $0.825 | $3.30 | $0.0035 | $0.3465 |
| DeepSeek V4.1 Flash 1017 Abliterated | deepseek-v4.1-flash-1017-abliterated | Abliterated | 1M tokens | $0.75 | $3.125 | $0.0032 | $0.3188 |
| DeepSeek V4.1 Flash 1017 Heretic | deepseek-v4.1-flash-1017-heretic-uncensored | Abliterated | 512K tokens | $0.825 | $3.30 | $0.0035 | $0.3465 |
| DeepSeek V4.1 Flash Abliterated | deepseek-v4.1-flash-abliterated | Abliterated | $0.8488 | $1.6975 | $0.0031 | $0.3056 | |
| DeepSeek V4.1 Flash 0910 Uncensored | deepseek-v4.1-flash-0910-uncensored | Uncensored | 524K tokens | $0.825 | $3.30 | $0.0035 | $0.3465 |
| DeepSeek V4 Pro Uncensored | deepseek-v4-pro-uncensored | Uncensored | $0.97 | $1.94 | $0.0035 | $0.3492 | |
| DeepSeek V4 Pro Abliterated | deepseek-v4-pro-abliterated | Abliterated | 1M tokens | $1.00 | $4.00 | $0.0042 | $0.42 |
| DeepSeek V4.1 Flash 1017 Uncensored Thinking | deepseek-v4.1-flash-1017-uncensored-thinking | Uncensored, thinks first | 1M tokens | $0.825 | $3.30 | $0.0035 | $0.3465 |
| DeepSeek V4.1 Flash 1017 Heretic Thinking | deepseek-v4.1-flash-1017-heretic-uncensored-thinking | Uncensored, thinks first | 512K tokens | $0.825 | $3.30 | $0.0035 | $0.3465 |
| DeepSeek V4.1 Flash Uncensored Thinking | deepseek-v4.1-flash-uncensored-thinking | Uncensored, thinks first | 1M tokens | $0.825 | $3.30 | $0.0035 | $0.3465 |
| DeepSeek V4.1 Flash 1017 | deepseek-v4.1-flash-1017 | Stock | 1M tokens | $0.25 | $1.20 | $0.0011 | $0.111 |
| DeepSeek V4.1 Flash | deepseek-v4.1-flash | Stock | 1M tokens | $0.06 | $0.18 | $0.00023 | $0.0234 |
| DeepSeek V4 Pro 0423 | deepseek-v4-pro | Stock | 1M tokens | $0.792 | $2.376 | $0.0031 | $0.3089 |
| DeepSeek V3.2 | deepseek-v3.2 | Stock | $1.08 | $1.6801 | $0.0037 | $0.3744 | |
| DeepSeek R1 | deepseek-r1 | Stock | $2.00 | $8.7202 | $0.0086 | $0.8616 |
USD per million tokens, read from the RelaySwitch catalog when this page was built. A message here is 3,000 tokens in (card, persona and history) and 300 out.
Set Context Size to the model's context window or less; SillyTavern trims the history to fit it, and a larger value sends more tokens per message. A build marked "thinks first" reasons before it answers and spends part of Max Response Length on that: raise it to 1,000 or more, or pick the same build without the thinking step. The best SillyTavern models list ranks every chat model on the key by use in the last seven days.
03One source for every DeepSeek build
DeepSeek's own API serves the stock models only. Through RelaySwitch one Custom source reaches the stock builds, the uncensored and abliterated builds, and every other model on the Models page, so switching between them is a change of Model ID, not a new connection profile. A failed request is handed to the next host for that model at the same price, and every reply carries a signed receipt you can check on the verify page.
To confirm the key sees the DeepSeek ids from a terminal:
curl https://api.relayswitch.io/v1/models -H "Authorization: Bearer sk-relay-..." | grep deepseek04Sources
Field names are SillyTavern's own, from the API Connections panel on the release branch as read for the SillyTavern settings guide on 2026-11-02. Model ids, tags, context windows and prices come from the RelaySwitch catalog at render. The same key runs DeepSeek on Janitor AI, Chub AI and RisuAI.
Questions
- How do I connect SillyTavern to the DeepSeek API?
- Set API to Chat Completion, Chat Completion Source to Custom (OpenAI-compatible), Custom Endpoint to https://api.relayswitch.io/v1, Custom API Key to a sk-relay- key, and Enter a Model ID to a DeepSeek id from the table. Click Connect.
- Which DeepSeek model is best for SillyTavern?
- Judge by the table: it prices the same message on every DeepSeek build on the key. DeepSeek V4.1 Flash 1017 Uncensored heads the uncensored list at $0.0035 for one 3K-in / 300-out message today; the stock builds cost less per message and keep DeepSeek's own refusals. The best SillyTavern models list ranks them by actual use.
- What Context Size should I set for DeepSeek?
- The context window in the table, or less. SillyTavern sends as much history as fits the setting, so a larger Context Size costs more per message once a chat has grown.
- Why are DeepSeek replies cut off or empty?
- A build that thinks first spends part of Max Response Length on its reasoning. Raise Max Response Length to 1,000 or more, or use the same build without the thinking step.
- Is there a free DeepSeek API for SillyTavern?
- Not here. Keys are prepaid from $10 with no subscription, and every message bills its tokens from that balance at the rates in the table.
- Can I use uncensored DeepSeek in SillyTavern?
- Yes. The builds the catalog tags uncensored or abliterated are in the table with their ids; type one as the Model ID on the same Custom source. The uncensored models page shows the daily check for each.