# Martian

> LLM model router and unified API that selects the best model per request to improve performance and reduce cost.

Martian provides a model router and gateway that forwards each request to the most suitable LLM based on performance, cost, and reliability. It’s a drop-in, OpenAI-compatible endpoint with controls for max cost and willingness-to-pay, automatic failover across providers, benchmarking reports, and access to a large catalog of supported models. A free developer tier includes 2,500 requests; paid usage is metered, and enterprises can deploy custom routers with SLA and VPC options.

## Features
- Real-time routing across multiple LLM providers
- OpenAI-compatible API (base_url swap) for easy integration
- Cost/performance controls (e.g., max_cost, willingness_to_pay)
- Automatic failover and outage rerouting for higher uptime
- Benchmarking reports to validate router performance
- Unified access to hundreds of models via adapters
- Single Martian key or bring-your-own provider keys
- Enterprise options: SLA, VPC deployment, and custom-built routers
- Supported SDK examples for Python, Node, and LangChain
- Leaderboards and datasets to compare model cost/latency/performance

## Integrations
OpenAI SDK (Python/Node), Anthropic, LangChain (Python/JS), DeepInfra, Cerebras

## Platforms
WEB, API, DEVELOPER_SDK

## Pricing
Open Source, Free tier available

## Links
- Website: https://www.withmartian.com
- Documentation: https://docs.withmartian.com/martian-model-router
- Repository: https://github.com/withmartian/llm-adapters
- EveryDev.ai: https://www.everydev.ai/tools/martian