Cheaper · 160 models · Invoices

See pricing
TVVV

API · OPENAI / ANTHROPIC

API

One key for Claude Opus 5.5 / Sonnet 5.5, GPT-6, DeepSeek, GLM, Kimi and free models over OpenAI- and Anthropic-compatible endpoints — Claude Code, Cursor, Cline and Cherry Studio work by changing the base URL. Seedance and MiniMax H3 video too. Shares your tvvv credit balance; billed by actual usage.

OPENAI

https://tvvv.ai/api/v1

base_url for OpenAI SDK, Cursor, Cline, Cherry Studio

ANTHROPIC

https://tvvv.ai/api

ANTHROPIC_BASE_URL for Claude Code and the Anthropic SDK

API keys

…

Sign in

Quick start

Run in a terminal (or put in ~/.claude/settings.json → env)
export ANTHROPIC_BASE_URL=https://tvvv.ai/api
export ANTHROPIC_AUTH_TOKEN=tvvv_sk_…
export ANTHROPIC_MODEL=claude-opus-5.5
export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4.5
claude

Use the full id (anthropic/claude-opus-5.5) or short forms like claude-opus-5.5 / claude-opus-5-5. GET /api/v1/models lists every model with prices.

Models and prices

Credits per 1M tokens (10 credits = ¥1), pay-as-you-go rate; member credits cost less. Billed by actual usage, at least 1 credit per request.

ModelInputCache readOutput
Claude Opus 5.5anthropic/claude-opus-5.5305151523
Claude Sonnet 5.5anthropic/claude-sonnet-5.51528762
Claude Haiku 4.5anthropic/claude-haiku-4.59510476
GPT-6 Astraopenai/gpt-6-astra952954760
DeepSeek V4 Flashdeepseek/deepseek-v4-flash-07317115
GLM-5.3z-ai/glm-5.388141
Kimi K3moonshotai/kimi-k324211199
Qwen3.8 Flashqwen/qwen3.8-flash14245
Gemini 3.8 Flashgoogle/gemini-3.8-flash717357
GLM-5.3 Flashz-ai/glm-5.3-flash15350
Ling 3.1 Flashinclusionai/ling-3.1-flash8223
DeepSeek V3.2deepseek/deepseek-v3.228342
GLM-5z-ai/glm-510020321
Kimi K2.6moonshotai/kimi-k2.66515343
Qwen3.6 Plusqwen/qwen3.6-plus33—196
North Mini CodeFreecohere/north-mini-code:freeNo token charge — 1 credit per request
Laguna S 2.1Freepoolside/laguna-s-2.1-freeNo token charge — 1 credit per request
Nemotron 3 UltraFreenvidia/nemotron-3-ultra-550b-a55b:freeNo token charge — 1 credit per request
Nemotron 3.5 LightningFreenvidia/nemotron-3.5-lightning:freeNo token charge — 1 credit per request
InklingFreethinkingmachines/inkling:freeNo token charge — 1 credit per request
Gemma 4 31BFreegoogle/gemma-4-31b-it:freeNo token charge — 1 credit per request
  • Before each request we hold the worst case (estimated input + max_tokens output); with too little balance you get 402 and no model is called. Afterwards you pay actual usage and the rest is released.
  • Max 32,000 output tokens per request (larger max_tokens is clamped). Use streaming for long outputs.
  • Streaming, tool calls, image input, Claude prompt caching and thinking are supported. Not supported yet: server-side web search / code execution tools, fast mode, n > 1.
  • Non-streaming responses carry an x-tvvv-credits header with the credits charged; usage history is on your account page.

Video: Seedance / H3

POST /api/v1/videos/generations returns a task id at once; poll GET /api/v1/videos/generations/{id} every 20–30 s until it returns a signed video_url (valid ~6 h). Videos are also saved to your library. The quoted price is held and only charged on success. Params: model, prompt, duration, resolution, aspect_ratio, first_frame_url, last_frame_url, reference_image_urls, generate_audio.

Seedance 2.5bytedance/seedance-2.5-fullfrom 30 credits
Seedance 2.5 Previewbytedance/seedance-2.5-previewfrom 30 credits
MiniMax H3minimax/hailuo-32K 12.4 / 480p 0.8 / 720p 1.2 / 1080p 1.5 credits per second
MiniMax H3 Maxminimax/hailuo-3-max480p 4.8 / 768p 7.7 credits per second
MiniMax H3 Max Turbominimax/hailuo-3-max-turbo480p 2.4 / 768p 3.9 / 1080p 7.7 credits per second
H3 Max 多机位minimax/h3-max-multi-angle480p 4.8 / 768p 7.7 / 1080p 15.3 credits per second
H3 Max 视频风格化minimax/h3-max-styles768p 7.7 credits per second
Seedance 2.0bytedance/seedance-2.0from 20 credits
Seedance 2.0 Minibytedance/seedance-2.0-minifrom 10 credits
Seedance 2.0 Fastbytedance/seedance-2.0-fastfrom 16 credits
Seedance 1.5 Probytedance/seedance-1-5-pro720p 2.5 / 1080p 5.6 credits per second; with audio 720p 5 / 1080p 11.2

Full video prices onrates; to let an agent generate images and video, useMCP.