LiteLLM overview and key facts (October 2026)

LiteLLM is an open-source Python SDK and self-hosted AI gateway that lets developers call more than 100 language models through one OpenAI-style interface. IT teams use the proxy to control keys, budgets and costs for all model access in the company.

The project consists of a Python library and a proxy server that supports providers such as OpenAI, Anthropic, Vertex AI and Bedrock. It is developed by BerriAI, a Y Combinator company. The GitHub repository had about 59,800 stars and a last push on September 28, 2026. An enterprise edition adds single sign-on, audit logs and support with response times.

• Switches models without code changesApplications call every model with the same completion interface and receive the same output format. Teams can change providers or test new models without rewriting integrations.
• Keeps services running during outagesRetries, fallbacks and load balancing route requests to another deployment or provider when one fails. Internal AI services stay available.
• Controls spending per teamVirtual keys carry budgets per key, team or user, and the proxy tracks costs. Finance and IT see who uses which model and can stop overruns.

• Runs entirely on your own infrastructure, including air-gapped deployments in the enterprise edition.
• The open-source core is MIT-licensed; enterprise features sit in a separately licensed directory.

LiteLLM is an open-source project by BerriAI and can be self-hosted worldwide without country ties.

You operate the gateway yourself, so plan hosting, updates and monitoring. Check which features you need from the enterprise directory, since single sign-on and audit logs are licensed separately. The gateway sees every prompt, so treat its logs as sensitive data.

No reviews yet. Be the first to share how your team uses it.

Every review shows its verification level. We read each one before it goes live. How we check reviews