OpenRoute is recruiting agents — share the AI gateway market →Bare-metal GPU rentals from 1 hour — idle-compute scheduling platform coming soon →
OROpenRoute
  • Models
  • Bare Metal
  • Observability
  • Docs
  • Blog
OROpenRoute· Ontology Route for AI Agent
ModelsProvidersObservabilityGuardrailsPricingBare MetalTerms of ServicePrivacy Policy

Prices are shown in CNY (¥) per million tokens and are for reference only. Live pricing, quotas and billing are subject to the console.

Copyright ©2025-2027 天津聚链科技有限公司版权所有

津ICP备2023007973号-5|beian津公网安备12011402001495号

One API, access all LLMs

One API, access all LLMs

Competitive pricing, stable uptime, unified model gateway + semantic guardrails + traffic insights for AI Agents.

Competitive pricing, stable uptime, unified model gateway + semantic guardrails + traffic insights for AI Agents.
app screenapp screen

Stop managing complexity, start building features

Three simple steps to quickly access leading AI models globally

Get StartedLearn More
无需信用卡,立即开始使用

FAQ

One API, One Bill, All major models
Try AI Image Apps
Dedicated GPU · Book bare metal
Get Started
Recruiting agents

OpenRoute is recruiting agents

OpenRoute is recruiting channel partners as agents for the unified multi-model AI gateway, with competitive partnership policies and dedicated support.

Learn about partnership

Popular models

Hand-picked trending models with catalog pricing.

All models

DeepSeek-V4-Flash

deepseek

对话

工具调用函数调用流式输出

DeepSeek-V4-Pro

deepseek

对话

长上下文推理能力代码生成

MiniMax-M2.7

minimax

对话

工具调用函数调用流式输出

mimo-v2.5-pro

xiaomi_mimo

对话

工具调用函数调用流式输出

kimi-k3

moonshot

对话

工具调用函数调用流式输出

glm-5.2

z-ai

对话

工具调用函数调用流式输出

Connected providers

One route to every inference platform, with automatic failover and load balancing.

All providers
Qwen

Qwen

24 models · 0

MiniMax

MiniMax

16 models · 0

Zai

Zai

12 models · 0

Deepseek

Deepseek

5 models · 0

VolcEngine

VolcEngine

5 models · 0

Dashscope

Dashscope

4 models · 0

Doubao

Doubao

4 models · 0

Moonshot

Moonshot

3 models · 0

BAAI

BAAI

2 models · 0

MIMO

MIMO

2 models · 0

All providers

Why Ontology Route for AI Agent

Ontology at the core — turning complex model orchestration into a reliable, observable and governable layer.

Transparent CNY pricing

All prices are listed in CNY per million tokens, with clear input, output and cache breakdowns.

Multi-provider aggregation

The same model may come from several providers — compare prices and fail over seamlessly.

Connectivity monitoring

Track latency, throughput and availability per route in real time; always route to the healthiest node.

Unified parameters

Standardized parameter constraints and defaults across vendors — one schema for chat, reasoning and multimodal.

All-modality coverage

Chat, reasoning, embeddings, image and audio through one API, filterable by type and capability.

Usage insights

Aggregate 30-day call volume by model and provider to spot trends and long-tail usage.

Find the right model<br/>and price for every call

OpenRoute describes model capabilities with an ontology (Ontology) and uses routing (Route) to deliver each request to the optimal provider. Language, audio, image, and video models share one interface with transparent billing.

Browse modelsLearn about guardrails
Language
5
LLM · Chat & reasoning
Audio
3
Recognition & synthesis
Image
3
Generation & understanding
Video
2
Generation & editing
Observability & Evaluation

Every call, clearly visible

Ontology Route for AI Agent ships with observability and evaluation built in. Without touching your code, capture the full call chain of every run, score output quality continuously, and make production behavior measurable, traceable and improvable.

  • End-to-end tracing

    From the entry request to every tool call, sub-task dispatch and model response — the full call tree and timing waterfall.

  • Automated evaluation

    Built-in faithfulness, relevancy, toxicity and more, with custom scoring rules and human annotation feedback.

  • Metrics dashboards

    Latency, tokens, cost and success rate aggregated by model, task and tenant, with anomaly alerts.

Explore observability
Trace#8f2a…c1
2.84s · 6 spans
agent.run
2.84s
planner · gpt-4o
640ms
tool: search_docs
510ms
retriever.embed
280ms
executor · claude-3.5
980ms
tool: run_code
390ms

Faithfulness

0.94

Relevancy

0.88

Safety

0.99

Supported Pricing Models

Billing naturally differs across modalities. OpenRoute annotates every model with its native pricing basis and converted price.

  • Per token
  • Per duration
  • Per item
  • Tiered
  • Subscription
View all 80 models & pricing
Bare-metal GPU rentals from 1 hour — idle-compute scheduling platform coming soon →

At scale, owning GPU capacity costs less

Bare-metal short-term and monthly rentals for heavy gateway users: dedicated hardware, on-demand delivery. An idle-compute scheduling platform (RunPod / Novita style) is coming soon — reserve now for early-bird pricing.

Book bare metalSee the roadmap

guardrails · observability

Guardrails & Full-Chain Observability

Guardrails enforce policies on both sides of every model call, while traces capture each span of agent planning, retrieval, tool calls and model inference. Call chains, cost and quality metrics are fully queryable with no changes to your code.

Guardrails Policies
  • PII Masking · Before input · Rewrite
  • Prompt Injection Detection · Before input · Block
  • Keyword Filter · Before input · Warn
  • Content Moderation · After output · Block
  • JSON Schema Validation · After output · Rewrite
Observability Metrics
P95 Latency
2.4s
End-to-end full-chain duration
Trace Retention
30d
Default retention period
Eval Pass Rate
98.2%
Sampled scoring from automated evaluators
Cost Visibility
100%
Every call billed by tokens and amount
Guardrails Pipeline
6 policies

Request guardrails · Input

PII detectionUnavailable · needs self-hosted servicePass
Jailbreak injectionBlock
KeywordsPass
model call

Response guardrails · Output

Content moderationPass
JSON SchemaPass
Hallucination checkWarn
1 request blockedjailbreak · blocked
Guardrails

Safety guardrails on every call

Ontology Route for AI Agent has Guardrails built into the gateway layer. Without touching your business code, validate content entering and leaving the model for compliance, safety and format — stopping risk before it reaches your users, with every block traced in the observability chain.

  • Bidirectional input / output checks

    Validate before the request reaches the model and before the response returns — block, rewrite or degrade on policy hit.

  • Structured output validation

    Enforce JSON Schema on model output and auto-retry on missing or invalid fields to keep downstream stable.

  • Policy as code

    Version-managed guardrail rules, mountable per task, tenant and environment, with every hit recorded in the observability chain.

Explore guardrails

Ready to connect 80+ models?

Start from the model catalog to find the best model and pricing by provider, type and context length.

Input

¥3.00

Output

¥9.00

Context

1M

by deepseekAvailable

Input

¥9.00

Output

¥27.00

Context

1M

by deepseekAvailable

Input

¥2.10

Output

¥8.40

Context

205K

by minimaxAvailable

Input

¥3.00

Output

¥6.00

Context

100K

by xiaomi_mimoAvailable

Input

¥20.00

Output

¥100.00

Context

1.048576M

by moonshotAvailable

Input

¥8.00

Output

¥28.00

Context

1M

by z-aiAvailable
View guardrails & observability details
Browse models
View providers