LiteLLM Cheatsheet

Q: How do I install and run LiteLLM?

Run `pip install litellm` then `litellm --model gpt-4o --port 4000`. Set `OPENAI_API_KEY` in your environment. Configure multiple models in a YAML config file and start with `litellm --config config.yaml`.

Q: How do I configure multiple models?

Create a YAML config with model_list containing model_name, litellm_params (model, api_key), and model_info. Start with `litellm --config config.yaml --port 4000`. The proxy routes requests based on the model field.

Q: Can LiteLLM track costs?

Yes, use `--max-budget` to set a monthly spending limit. LiteLLM tracks token usage and cost per model. It can also enforce spending limits per API key or user via the config file.

LiteLLM is a lightweight LLM API proxy providing a unified OpenAI-compatible interface for 100+ models. Features cost tracking, rate limiting, model fallback, and load balancing — the standard solution for managing multi-model API calls.

Updated: 2026-07-20·12 commands

Quick Start

``bash # Install pip install litellm

# Start proxy for a single model litellm --model gpt-4o --port 4000

# Test with curl curl -s http://localhost:4000/v1/chat/completions \ -d '{"model":"gpt-4o","messages":[{"role":"user","content":"Hello"}]}'

# List available models curl -s http://localhost:4000/v1/models | jq .`

`Single Model Mode`

Quickest way: litellm --model . Supports all OpenAI-compatible models plus Claude, Gemini, Llama, Mistral, and 100+ others through provider-specific configuration.

`Multi-Model Config`

For production, use a YAML config file defining multiple models with different providers, rate limits, and cost settings. Each model appears as a separate entry in the /v1/models endpoint.

`Cost Management`

--max-budget` sets a hard monthly cap. LiteLLM logs every request's model, tokens, and cost. The config file supports per-user and per-key spending limits for multi-tenant setups.

Startup & Modes(2)

Command	Level	Example
`litellm --help` Show LiteLLM help	Basic	`litellm --help`
`pip install litellm` Install LiteLLM	Basic	`pip install litellm`

ai-server(10)

Command	Level	Example
`litellm --model` Specify model and start proxy (core usage)	Basic	`litellm --model gpt-4o --port 4000`
`litellm --model --port` Specify proxy port	Basic	`litellm --model claude-3-opus --port 4000`
`litellm --config` Start proxy with config file	Intermediate	`litellm --config ./litellm_config.yaml --port 4000`
`litellm --drop-params` Drop sensitive params from requests	Intermediate	`litellm --model gpt-4o --drop-params api_key`
`litellm --max-budget` Set monthly cost limit	Intermediate	`litellm --model gpt-4o --max-budget 50.0`
`curl localhost:4000/v1/chat/completions` Send OpenAI-compatible request through proxy	Intermediate	`curl -s http://localhost:4000/v1/chat/completions -d '{"model":"gpt-4o","messages":[{"role":"user","content":"Hello"}]}'`
`curl localhost:4000/v1/models` List available models through proxy	Basic	`curl -s http://localhost:4000/v1/models \| jq .`
`curl localhost:4000/health` Check proxy health	Basic	`curl -s http://localhost:4000/health`
`litellm --num_requests` Set max concurrent requests	Intermediate	`litellm --model gpt-4o --num_requests 10`
`litellm --request_timeout` Set request timeout (seconds)	Intermediate	`litellm --model gpt-4o --request_timeout 600`

FAQ

How do I install and run LiteLLM?

How do I configure multiple models?

Can LiteLLM track costs?

This cheatsheet is compiled from official tool documentation. Last updated: 2026-07-20.