XinoAPI is a Chinese AI API gateway for language and video-generation models including DeepSeek, Qwen, GLM, Kimi, MiniMax, Wan, and ByteDance Seed.

XinoAPI is intended for developers and businesses outside mainland China who need Chinese language and video models with Stripe billing and unified routing.

Does XinoAPI store prompts and responses?

XinoAPI does not retain plaintext prompts or responses by default. Billing and abuse-prevention systems use metadata such as model, token counts, status, latency, and timestamps.

XinoAPI: Chinese Language & Video Models Through One API

// Quickstart

One line changes.
Everything else stays.

Keep your SDK

Works with the official OpenAI client in Python, Node, and every compatible toolchain.

Explicit model IDs

model=deepseek-v4-flash means exactly that. No silent rerouting.

$2.00 free credits

Self-service onboarding with Stripe billing. No China entity, no Alipay, no ICP.

quickstart.py

# Just change one line. That's it.
 
from openai import OpenAI
 
client = OpenAI(
  api_key="xino-your-key-here",
  base_url="https://api.xinoapi.com/v1"  # ← only this line
)
 
response = client.chat.completions.create(
  model="deepseek-v4-flash",
  messages=[{"role": "user", "content": "Hello!"}]
)
 
print(response.choices[0].message.content)

// Why XinoAPI

Built for global developers

Singapore, Hong Kong, and Tokyo relay paths with dedicated enterprise routes. Just Stripe, one API key, and every frontier model.

<200ms

Optimized Global Routing

Requests ride the lowest-latency path into each upstream, with smart failover across channels when a route degrades.

Zero

Prompt Storage

No plaintext prompts or completions retained. Billing uses metadata only. Optional local redaction via Privacy SDK.

Explicit

Model IDs, No Black Box

What you request is what you get. No mystery model swaps — every response carries the exact upstream model identifier.

// Models

Leading Chinese AI models,
one account.

OpenAI-compatible language routes plus separately metered asynchronous video generation.

Model	Context	Input	Output	Latency
DeepSeek V4-Flash SAVE 10%	1M	$0.126/M*	$0.252/M*
DeepSeek V4-Pro SAVE 10%	1M	$1.521/M*	$3.042/M*
Qwen Plus NEW	1M	$0.133/M	$0.800/M
Qwen Max FLAGSHIP	256K	$0.400/M	$1.600/M
GLM-5.3 LATEST	1M	$1.192/M	$4.172/M	—
GLM-5.2 SAVE 10%	1M	$1.260/M*	$3.960/M*
Kimi K3 NEW	1M	$3.600/M	$18.00/M
Kimi K2.6 STABLE	256K	$1.083/M	$4.500/M
MiniMax M3 NEW	1M	$0.350/M	$1.400/M
Dola Seed 2.1 Turbo NEW	256K	$0.500/M	$2.500/M	—

* Featured prices are 10% below the matching Baidu Qianfan International list rate; cache-hit input is lower. Rates verified September 17, 2026. View full pricing →

// Video generation

Three current video families, one published rate card.

Native provider meters stay visible, so seconds and generation tokens are never mixed.

Model	Billing	Starting rate	Maximum output	Highlights
MiniMax H3 NEW	Per second	$0.080/sec	2K · 15 sec	Native stereo audio
Wan 3.0 Video NEW	Per second	$0.056/sec	1080P · 30 sec	Generation + editing
Seedance 2.5 NEW	Generation tokens	$6.40/M*	720P · 30 sec	Up to 50 references

* Seedance 2.5 is $6.40/M generation tokens with video input and $10.70/M without video input. See video pricing details →

// Get Started

Start building with
Chinese frontier models.

Self-service onboarding, $2.00 free credits, and 6 provider families across language and video routes.

Create account Sign in

Self-service · Supported regions only · Privacy SDK available

One line changes.Everything else stays.