OpenRouter教程：一个API Key用所有LLM
不用再为每个模型服务商维护单独的API Key。
OpenRouter Tutorial 2026: One API Key for Every LLM
OpenRouter is an API gateway that gives you access to hundreds of models - Anthropic, OpenAI, Google, DeepSeek, Qwen, Llama, and more - through a single OpenAI-compatible endpoint and one API key. In 2026 it is the standard way developers compare and route models without managing a dozen accounts.
Why Developers Use It
- One key, one SDK - swap models by changing one string
- Pay-as-you-go - credits work across all providers
- Model comparison - test the same prompt on 10 models in minutes
- Fallback routing - if one provider is down, auto-failover to another
- No minimums - start with $5
Getting Started
# 1. Sign up at openrouter.ai, create a key (sk-or-v1-...)
# 2. Standard OpenAI SDK
from openai import OpenAI
client = OpenAI(
api_key="sk-or-v1-...",
base_url="https://openrouter.ai/api/v1"
)
resp = client.chat.completions.create(
model="anthropic/claude-sonnet-4",
messages=[{"role": "user", "content": "Hello!"}]
)
print(resp.choices[0].message.content)
Model IDs You Should Know
| Model | OpenRouter ID | Cost class |
|---|---|---|
| Claude Sonnet 4 | anthropic/claude-sonnet-4 | Mid |
| GPT-5 mini | openai/gpt-5-mini | Low |
| DeepSeek V3 | deepseek/deepseek-chat | Very low |
| Qwen 2.5 72B | qwen/qwen-2.5-72b-instruct | Low |
| Llama 3.3 70B | meta-llama/llama-3.3-70b-instruct | Low |
| Free models | various (:-free suffix) | $0 |
Advanced Features
Automatic routing:
"model": "openrouter/auto"
Routes each request to the best model for the prompt - useful for cost optimization.
Fallbacks:
"model": ["anthropic/claude-sonnet-4", "openai/gpt-5-mini"]
If the first fails, the second is used.
Cost tracking: The response includes usage.cost - log it to see what you spend per request.
FAQ
Is OpenRouter cheaper than direct APIs? Slightly more per token (small gateway fee) but often cheaper in practice because you can pick the cheapest adequate model per task.
Is my data private? OpenRouter is a pass-through; your data goes to the model provider. Enterprise tier offers zero-retention with some providers.
Free models on OpenRouter? Yes - several models are available free (rate-limited) for experimentation.
