LiteLLM
Open-source, self-hosted gateway to 100+ LLM APIs in the OpenAI format.
A Python SDK and proxy server that calls 100+ model providers through one OpenAI-compatible API, with load balancing, fallbacks, spend tracking, budgets and virtual keys. The most common self-hosted choice for running a gateway inside your own infrastructure.
- Vendor
- BerriAI
- Category
- AI gateways
- Pricing
- Open source
- Open source
- Yes
- Platforms
- Python, Docker, Self-hosted
- Launched
- 2023
- Website
- litellm.ai
Features
- OpenAI-compatible proxy
- Fallbacks & load balancing
- Budgets & virtual keys
Best for
- Enterprise
- High volume / low cost
More ai gateways
- OpenRouter — OpenRouter. One API for hundreds of models.
- Portkey — Portkey. Production AI gateway with routing, guardrails, budgets and observability.
- Cloudflare AI Gateway — Cloudflare. Edge gateway for caching, rate-limiting and observing AI API calls.
- Vercel AI Gateway — Vercel. One endpoint for hundreds of models, with no markup on tokens.
- Azure API Management AI gateway — Microsoft. AI gateway policies built into Azure API Management.
- Kong AI Gateway — Kong. AI plugins that turn the Kong API gateway into a multi-LLM gateway.
- Apigee AI gateway — Google Cloud. AI gateway policies in Google Cloud’s Apigee API platform.
- Bifrost — Maxim AI. Fast open-source gateway in Go with budgets, virtual keys and an MCP gateway.