OpenRouter is a unified API that gives you access to 300+ AI models through a single endpoint. Instead of managing separate accounts with Anthropic, OpenAI, Google, and others, you use one API key and one credit balance.
How it works
- You send a request to OpenRouter’s API
- You specify which model you want (e.g.,
anthropic/claude-opus-4.6) - OpenRouter routes it to the right provider
- You get the response in a standard format
The API is OpenAI-compatible, so any code that works with OpenAI works with OpenRouter — just change the base URL.
Why developers use it
- One API key for Claude, GPT, Gemini, DeepSeek, Qwen, Kimi, and 300+ more
- 25+ free models — including Qwen 3.6 Plus and DeepSeek V3
- Fallback routing — if one provider is down, auto-switch to another
- Works with every tool — Aider, OpenCode, Continue.dev
Fallback routing explained
One of OpenRouter’s most valuable features is automatic fallback. If you request Claude and Anthropic’s API is experiencing issues, OpenRouter can automatically route to a backup provider or model you’ve specified. This is critical for production applications where downtime means lost revenue.
You can configure fallback chains:
{
"route": "fallback",
"models": [
"anthropic/claude-sonnet-4.6",
"openai/gpt-5.2",
"deepseek/deepseek-chat"
]
}Free models
OpenRouter offers 25+ models at zero cost, subsidized by the platform. These include:
- Qwen 3.6 Plus (Alibaba’s flagship)
- DeepSeek V3 (strong coding model)
- Gemma 4 9B (Google’s open model)
- Various Llama and Mistral variants
Free models have rate limits but are perfect for development, testing, and personal projects.
Pricing
OpenRouter charges a 5.5% fee on credit purchases. After that, you pay each model’s rate. Models range from free to $15/1M tokens (Claude Opus).
| Model | Input (per 1M) | Output (per 1M) |
|---|---|---|
| DeepSeek Chat | $0.27 | $1.10 |
| MiniMax M2.7 | $0.30 | $1.20 |
| Claude Sonnet 4.6 | $3.00 | $15.00 |
| Claude Opus 4.6 | $15.00 | $75.00 |
| GPT-5.2 | $2.50 | $10.00 |
Quick start
from openai import OpenAI
client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key="your-key")
response = client.chat.completions.create(
model="deepseek/deepseek-chat", # any of 300+ models
messages=[{"role": "user", "content": "Hello"}]
)Using OpenRouter with coding tools
Most AI coding tools support OpenRouter natively:
# Aider
aider --model openrouter/anthropic/claude-sonnet-4.6
# OpenCode
opencode --provider openrouter --model deepseek/deepseek-chatFor Continue.dev, add OpenRouter as a provider in your config and access any model through it.
FAQ
Is OpenRouter just a reseller, or does it add value?
OpenRouter adds significant value beyond reselling. It provides unified billing across all providers, automatic fallback routing for reliability, rate limit management, usage analytics, and a standardized API format. The 5.5% fee covers these features — many developers find it cheaper than managing multiple provider accounts separately.
Are there any downsides to using OpenRouter instead of direct APIs?
The main trade-offs are slightly higher latency (requests route through an extra hop) and the 5.5% markup on credits. For latency-sensitive applications like real-time autocomplete, direct API access may be preferable. For most other use cases, the convenience and reliability features outweigh the small overhead.
Can I use OpenRouter for production applications?
Yes. OpenRouter is used in production by many companies and tools. Its fallback routing actually makes it more reliable than direct API access for production use, since it can automatically switch providers during outages. They offer usage-based pricing with no minimum commitments.
Learn more
- OpenRouter Complete Guide — full setup, pricing, and model selection
- OpenRouter vs Direct API — when to use each
- Best Free AI APIs 2026 — all free options compared
Related: Ai Glossary Developers