AI Unified via Gateway

One Key, Unified Access
Enterprise AI Model Gateway Routing

An OpenAI-compatible multi-model scheduling gateway for duly registered, compliant models in mainland China such as DeepSeek, Kimi and Qwen — one-stop auth, rate limiting, routing, failover, billing and log management. Automatic multi-upstream failover, streaming responses and token-level usage stats; no code changes needed, just swap the Base URL. The gateway does not modify models themselves and only routes duly registered models.

10+
Major models aggregated
99.9%
Multi-channel availability
<5ms
Gateway overhead
$0
Free trial credit on signup
chat-completions.py
# Change one line and you are on Aiuvia
from openai import OpenAI

client = OpenAI(
    api_key="sk-aiuvia-your-key",
    base_url="https://api.aiuvia.com/v1"
)

resp = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Hello"}]
)
print(resp.choices[0].message.content)
FEATURES

A gateway kernel built for production

From auth and rate limiting to routing and billing — engineering-grade, ready out of the box

OpenAI-compatible

Fully compatible with /v1/chat/completions and /v1/models. Plug in any OpenAI SDK, LangChain or Dify with a one-line Base URL change.

Smart multi-channel routing

Multiple upstream channels per model with priority and weighted load balancing. Automatic circuit breaking and second-level failover keep your business running.

Streaming SSE passthrough

Line-by-line flushed streaming with ultra-low first-token latency. Both streaming and non-streaming are billed on real token usage as they generate.

Token-level metered billing

Input and output are priced separately with flexible model multipliers. Balance, transactions and bills are visible in real time. Top up and go — no minimum spend.

Key-level security & rate limits

Independent RPM/TPM limits and daily quotas per API key, with IP allowlists and model allowlists. Abnormal calls trigger real-time alerts.

Full-chain call logs

Every call's model, channel, tokens, latency and status are fully logged. Customer-side detail and admin-side global views make issues instantly traceable.

MODELS

All major models in one place

Chat, reasoning and multimodal on demand. The catalog keeps growing — new models go live as soon as they ship

View all models & pricing →

DeepSeek-V3 / R1
Flagship · Cost-effective
GPT-4o / 4o-mini
OpenAI family
Kimi-K2
Long-context expert
Claude 3.5 Sonnet
Anthropic
Gemini 1.5 Pro
Google multimodal
GLM-4-Plus
Zhipu AI
Qwen-Max
Alibaba Qwen
+N
Custom channels
Any OpenAI-compatible upstream
ARCHITECTURE

One step. Fully managed.

Just swap the endpoint and your API Key — no code changes.

Your App

Zero code changes — call directly with OpenAI SDK / HTTP

Aiuvia Gateway

Auth · Rate limit · Routing · Failover · Billing · Logs

Model Upstreams

DeepSeek / OpenAI / Kimi / Claude / Gemini…

PRICING

Three plans that scale with you

Transparent pay-as-you-go pricing. Top up and use — balance never expires; the more you use per month, the better the unit price

Starter

Under $5,000/month · Sign up & use
Pay as you go
Same price as official · zero markup
  • All models, metered billing
  • Standard RPM / TPM limits
  • Streaming + non-streaming
  • Detailed call logs
Sign Up Free
Most Popular

Pro

From $5,000/month · prepaid
Metered top up and go
Input/output priced separately, real-time settlement
  • Everything in Starter
  • Higher RPM / TPM quotas
  • Multi-key group management
  • IP allowlist · model allowlist
  • Priority channel guarantee
Top Up Now

Enterprise

From $50,000/month · prepaid · contract pricing
0.X official base price
Tiered contract pricing by monthly volume — the more you use, the bigger the discount
  • Prepaid volume · contract-locked rates
  • Dedicated account pool, data isolation
  • Per-department sub-accounting
  • Corporate settlement · compliant invoicing
  • Embedded in finance / IAM / security workflows
Get a Quote
Billing notes: unified pricing with only two rates — input / output (per million tokens). Prompt tokens served from upstream caches are still billed at the regular input rate (caching is an internal cost-optimization and does not change prices). All prices are settled in real time against actual token usage.
QUICK START

Migrate in three lines of code

OpenAI SDK compatible — Python, Node.js and curl all work directly

Python
Node.js
curl
# pip install openai
from openai import OpenAI

client = OpenAI(
    api_key="sk-aiuvia-your-key",
    base_url="https://api.aiuvia.com/v1"  # your gateway URL
)
resp = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True  # streaming supported
)

Full API docs: sign in to the Console → “API Docs”

FAQ

Questions you may have

Do I need to change my code?▼
No. Aiuvia is fully compatible with the OpenAI API protocol. Just replace the Base URL in your existing code with the Aiuvia gateway URL and swap your API key. No SDK or framework changes.
Which models are supported?▼
DeepSeek, OpenAI GPT, Kimi, Claude, Gemini, GLM, Qwen and more, plus any OpenAI-compatible custom upstream. The live model list is visible in the Console “Model Plaza”.
What if a channel goes down?▼
Aiuvia has built-in health checks and circuit breaking: one model can have multiple upstream channels with weighted distribution. If a channel fails repeatedly, it is automatically isolated and traffic moves to backups — seamless failover.
How does billing work?▼
Token-level metered billing: input and output tokens are priced separately per model, and streaming requests are settled on actual usage. Balance is deducted in real time, every transaction is visible, no minimum spend, and balance never expires.
Can it be self-hosted?▼
Yes. Aiuvia ships a one-click deployment (single binary + install script + Nginx config) that runs on your own servers. Data never leaves your domain — ideal for enterprise AI platforms.

Ready to get started?

Get free trial credit on signup and go live in one minute — put every LLM to work for you

Sign Up Free