LiteLLM is an open-source AI gateway that lets developers call 140+ LLM providers (OpenAI, Anthropic, Google, AWS Bedrock, Azure, and more) through a single OpenAI-compatible interface. The self-hosted proxy adds virtual key management, per-user/project budget tracking, load balancing, fallback routing, rate limiting, and observability integrations with Langfuse, LangSmith, and OpenTelemetry. Enterprise features include SSO, RBAC, and team-level budget enforcement. As of mid-2026 it has approximately 40k GitHub stars. Key features: - Unified OpenAI-compatible API for 140+ LLM providers - Virtual API key management with per-user and per-project budget tracking - Load balancing, fallback routing, and rate limiting (RPM/TPM) - Observability integrations: Langfuse, LangSmith, OpenTelemetry - Prompt caching and 50% cost Batch API discounts - Enterprise tier: SSO, RBAC, and team-level spend enforcement
Open source (self-hosted): free. Enterprise: starts around $250/month; scales to approximately $30,000/year for premium tiers.