Aggregating China's AI Power · Serving Global Developers

One Gateway.
All Models.

80+ providers · 500+ models · Unified OpenAI-compatible API · Smart caching saves up to 90% tokens · Smart routing · Auto-failover · Pay-as-you-go

300T+tokens/mo
↓90%token cost savable
10M+users
99.4%uptime
Read Docs

Trusted by 10,000+ developers · 300T+ tokens monthly

Scroll down
500+
Models
80+
Providers
8ms
Fallback
99.4%
Uptime

Core Capabilities

Why Ai Token

Smart Token Savings · Up to 90% Off

We auto-inject prompt-cache breakpoints for Claude, GPT and more, so cached long contexts are billed at ~10% of the input price. Identical requests are served from our platform cache with zero upstream calls, and a max-tokens guardrail stops runaway agents from burning your balance. Every saving is shown live in your console.

How it saves

Smart Routing · Selectable Strategies · Auto-Failover

Switch between cost-first, latency-first and balanced routing on demand. We steer around a provider before it rate-limits or exhausts its quota, keep multi-turn chats pinned to one account, and fail over in 8ms — invisible to users.

Learn about routing

Usage-Based Pricing · Real-time & Transparent

Text by token, video and audio by the second, images by size tier — pay only for what you actually use, and see every tier's price before you call.

  • Text by token: input and output priced separately, with extra discounts on cache hits
  • Video & audio by the second: 5 seconds means 5 seconds — you never pay for unused duration
  • Images by size tier: 512 / 1K / 2K / 4K priced openly, larger sizes cost more
View pricing

Developer-First · Drop-in Replacement

Fully OpenAI SDK compatible. Python, Node.js, Go — just swap the base_url and you're done. Zero learning curve.

Read docs

Multi-Modal · One Key, Unified

Text, image, video, audio and embeddings — all orchestrated through a single API key, with no need to maintain multiple sets of credentials.

Browse models

Private Deployment

Keep your data on-premise: one-click Docker/K8s delivery, powered by the same AiToken gateway, ready for production traffic.

Explore enterprise

Auto-Hunt Free Tokens · Free on Signup

Stop hunting for free API keys yourself — we auto-discover, auto-connect and auto-cover them for you. Sign up and get platform-paid free credits instantly: cheap models like Claude Haiku and Gemini Flash cost you nothing, not a cent charged.

Claim free credits

Agent Routing · Exposed as an API

More than relaying tokens — the decision engine is exposed as /v1/route-decision: every request is routed in a structured, explainable way, with confidence, a fallback chain and a decision-model risk check. Pro / Enterprise only; Free gets basic relay.

See billing tiers

What Is AiToken – A Unified AI API Gateway?

AiToken is an enterprise-grade AI model aggregation gateway. It unifies DeepSeek, GPT, Claude, Gemini, Doubao, Kimi, Qwen, Midjourney, Sora and Veo behind one OpenAI-compatible /v1 endpoint: a single API key lets you switch freely between text, image, video, audio and embeddings—no separate sign-ups, integrations or SDKs per vendor. Multi-strategy smart routing fails over at the request level, steers around providers before they rate-limit or exhaust their quota, and degrades seamlessly to backup nodes when an upstream wobbles; billing is transparent and metered in CNY credits, and the cost calculator estimates your monthly spend up front. Whether you are an individual prototyping or an enterprise self-hosting (Docker / Kubernetes), AiToken covers the full path from a free key to production.

Decision Routing — the brain behind every request

AiToken does more than relay tokens. Its routing engine decides, per request, which upstream account to use — balancing cost, latency, quota and compliance. That is the difference between a dumb pipe and a managed gateway.

Billing tiers

Basic Relay

Included for all

Transparent smart routing under the hood — cost/latency/health balanced per request, with automatic failover.

Agent Routing

Pro / Enterprise

The decision engine exposed as an API: structured, explainable routing with confidence, fallback chain and a Bivia (Articul8) decision-model risk check.

Smart channel selection

Every request is decided to the optimal upstream account — by cost, latency, health and live rate-limit windows.

Automatic failover

When an upstream wobbles or hits quota, routing seamlessly degrades to backup nodes without breaking the call.

Geo-aware routing

Routes are steered by the real measured exit country, so geo-tagged fingerprints always land where they should — with geo-fence classification powered by the Articul8 Bivia decision model.

Explainable decisions

Each decision ships with confidence, the fallback chain and the live signals behind it — no black box.

Models

Models change; integration shouldn't

GPT-5.6 Sol
OpenAI
+2%$0.010/1K
DeepSeek V3
DeepSeek
+18%$0.001/1K
Claude 4
Anthropic
+5%$0.003/1K
Gemini 3.7 Flash
Google
+40%$0.0005/1K
Qwen3-235B
Alibaba
+12%$0.0008/1K
500+ Models80+ Providers25ms avg latency$0.0015 avg cost/1K tokens

Trending video models

View all
Seedance 1.0Bytedance
MiniMax-H3MiniMax
Kling 3.0Kuaishou
Seed 2.0 MiniBytedance

Trending image models

View all
Qwen-Image
Qwen-ImageAlibaba
Qwen-Image-Edit
Qwen-Image-EditAlibaba
Seedream 3.0
Seedream 3.0Bytedance
GPT Image 2
GPT Image 2OpenAI

Why AiToken

Stability and Compliance, Out in the Open

Others say "aggregation". We lay out every request's routing, fallback and compliance tier for you to see — proven by response headers, not by marketing.

Per-Request Observability

Every response carries X-Upstream-Channel / X-Degraded / X-Fallback-Hops headers. Which channel was hit, whether it degraded, how many hops — all visible in console and headers.

Native Multi-Protocol

OpenAI-compatible / native Anthropic / Gemini endpoints coexist. Claude Code, Cursor, Cline, Codex just point their Base URL at us — no self-built bridge.

Compliance-Tier Transparency

X-Upstream-Tier exposes exactly which compliance tier served the request; shadow-traffic gray launches are auditable — a layer freellmapi simply lacks.

MCP Catalog

A built-in MCP catalog lets agents discover and call your models and tools, turning the gateway into an orchestratable capability base.

One-Click Onboarding · 30s

Pick your tool → copy config → go live. No long docs: Base URL, Key and Model handed to you in one shot.

Start onboarding

Ready to get started?

Try Ai Token AI Gateway for free — simplify multi-model integration and unlock low-latency global access.

Contact Us