Connect Claude, Gemini, and DeepSeek through one OpenAI-compatible URL—with Claude and Gemini priced 20%–40% below official rates. Import our TavernAPI preset into SillyTavern to get tuned storytelling rules and local toggles, or connect directly with your own setup.
Verify your email (or sign in with Google or Discord) to get $0.30 free trial credit on Gemini 4 Argon and Gemini 3.8 Flash. Non-expiring balance or 30-day credit packs.
You always select the standard model ID in SillyTavern (like gemini-4-argon or claude-opus-5.5). Import our preset when you want our tuned storytelling rules, or use the raw model with your own local preset.
gemini-4-argon Connect https://tavernapi.com/v1 and pick any model ID from the dropdown. If you don't import a TavernAPI preset, we forward your prompt straight to the official 1M-context model with automatic prompt caching.
[[tav:pack/block]] Download our preset JSON from the console and import it into SillyTavern. You freely toggle word count, output language, and anti-impersonation rules locally, while locked core blocks ([[tav:pack/block]]) expand on our server.
Every model runs on an official upstream (Claude via AWS/GCP, Gemini via Google AI Studio, DeepSeek via the official API) with a full 1M context window. We skip GPT models because their roleplay prose feels stiff and repetitive. See our full hands-on rankings on the Models page.
#1 in overall roleplay experience (30% off official rates). Unmatched scene logic, character psychology, and strict adherence to your setting.
20% off official rates. Available if you personally like Sonnet, though we recommend most players pick Gemini 4 Argon or Claude Opus 5.5 instead.
Our #1 choice for most players (40% off official rates). The most natural, vivid phrasing and broadest world knowledge of any model we offer.
Lively Gemini prose at ~35% off official rates. Near-instant replies for high-volume daily chatting and testing new character cards.
Official DeepSeek API at one flat rate all day. Great value with sharp, expressive prose in English and Chinese when paired with our preset.
In a 32k-token roleplay chat, almost your entire bill is re-sending the chat history and character card on every reply. We place cache breakpoints on the server automatically—even over SillyTavern's standard OpenAI-compatible connection—and show your exact dollar savings in the console.
| Model | 50 replies without caching | 50 replies with TavernAPI caching |
|---|---|---|
| Claude Opus 5.5 | ~$4.76 | ~$1.03 ~78% less |
| Claude Sonnet 5.5 | ~$2.72 | ~$0.59 ~78% less |
| Gemini 4 Argon | ~$2.04 | ~$0.43 ~79% less |
In SillyTavern, click the plug icon at the top, set API to Chat Completion, set Chat Completion Source to Custom (OpenAI-compatible), and paste https://tavernapi.com/v1 and your sk-tav-... key.
Download our preset JSON from the console and import it in SillyTavern's AI Response Configuration panel (the sliders icon). Toggle word count, language, and anti-impersonation rules however you like.
Select gemini-4-argon, claude-opus-5.5, or any other model from the dropdown. Our server expands the locked [[tav:pack/block]] preset entries and caches your prompt automatically.
Your messages, character cards, and lorebooks are never saved to our database. We log only token counts, model name, timestamp, and cost per request for billing.
Pay as you go with USD balance that never expires, or grab a manual 30-day credit pack ($9 for $10, $19 for $23, or $39 for $48) when you want extra credit with no auto-renewal.
Claude and Gemini cost 20%–40% less than official rates, DeepSeek V4 Pro runs at one flat rate all day, and we never charge a 5.5% top-up fee.
We don't promise 'uncensored' magic—requests still pass through upstream providers' safety filters. However, we never ban your account just because an upstream model refused a prompt.
Download a TavernAPI preset JSON from the Presets page in our console, import it into SillyTavern's AI Response Configuration panel, and select a standard model ID like gemini-4-argon or claude-opus-5.5. You can freely toggle or edit plain-text rules (word count, output language, anti-impersonation) in SillyTavern, while locked core entries ([[tav:pack/block]]) expand automatically on our server. No plugins are needed, and you can also skip the preset to use your own local setup.
Your first top-up of $5 or more permanently unlocks the Standard preset. The Pro preset can be unlocked in three ways: buy it standalone for $4.99, use it for free while an active $19 or $39 30-day credit pack is in effect, or unlock it permanently once your cumulative spend reaches $25.
You receive $0.30 in free trial credit automatically after signing in and verifying your email address (or signing in with a verified Google or Discord account). Trial credit works right away on Gemini 3.8 Flash (gemini-3.8-flash) and Gemini 4 Argon (gemini-4-argon).
No, TavernAPI does not store your message content, character cards, or model replies. Our database records only the timestamp, model name, input/cached/output token counts, latency, and cost for each request so you can check your usage.
We intentionally exclude GPT models because they tend to produce stiff, repetitive, and overly moralizing prose in roleplay. We focus strictly on the five official 1M-context models that roleplay communities actually prefer: Claude Opus 5.5, Gemini 4 Argon, Gemini 3.8 Flash, DeepSeek V4 Pro, and Claude Sonnet 5.5.