BerriAI/litellm
★ 61,577 stars
Covered in week 2026-W21 · the 2026-05 report
Python SDK, Proxy Server (AI Gateway) to call 100+ LLM APIs in OpenAI (or native) format, with cost tracking, guardrails, loadbalancing and logging. [Bedrock, Azure, OpenAI, VertexAI, Cohere, Anthropic, Sagemaker, HuggingFace, VLLM, NVIDIA NIM]
What litellm is
LiteLLM is an open source AI Gateway that gives you a single, unified interface to call 100+ LLM providers — OpenAI, Anthropic, Gemini, Bedrock, Azure, and more — using the OpenAI format.
From the repository README · read it on GitHub →
Transcript
Portkey charges $99 a month for an LLM gateway. LiteLLM does it free. With over 47,000 stars on GitHub, LiteLLM is the open-source gateway that lets you call 100-plus LLM providers through a single OpenAI-compatible API. One endpoint, every model from GPT-4 to Claude to Gemini, with built-in load balancing, fallbacks, and spending limits across all of them. No more rewriting code every time you switch providers. Star it if this saved you money.