LiteLLM is a gateway you run yourself, with one OpenAI-compatible endpoint for 100+ providers. Makas is hosted for you and moves eligible requests to a cheaper model from the same provider.
You want lower model costs without running a gateway yourself.
Your key, your SDKYour provider key is sent with each request and not stored by Makas. To stop using Makas, point the base URL back at your provider.
Nothing to runNo gateway to deploy, update or scale. Change the base URL and add your Makas key.
Cheaper models, only when they suitBuilt-in strategies, on by default, move a request to a cheaper model from the same provider only when it suits the task. Each routed request is logged with both prices.
You want to run your own gateway with one OpenAI-compatible endpoint for 100+ providers.1
Virtual keysVirtual keys, each with its own models, budgets and rate limits.1
Load balancingLoad balancing, fallbacks and retries between deployments.1
Spend trackingTracks spend for keys, users and teams across 100+ LLMs.2
Side by side
Six things compared.
Makas and LiteLLM compared on six points
Makas
LiteLLM
Providers and models
OpenAI, Anthropic, GoogleWe add each provider’s new frontier models to the catalog as they are released. Routing stays within the provider your request arrived on, so there are no surprises.
100+ providersOne OpenAI-compatible endpoint for 100+ LLM providers.1
Provider keys
Not stored, passed throughYour own provider key, sent with each request and not stored by Makas.
Added to your gateway, encryptedYour own, added to the gateway you run and encrypted with your salt key.3
Who bills you
Your provider bills youYour provider, on your own provider account. Makas access is free on the Founding Customer plan. Planned prices
Free to self-hostSelf-hosting the open-source gateway is free; Enterprise is priced by gateway scale, not per token.4
Model choice
Automatic, on by default, strategies you switchEach request is classified by task and complexity, and moves to a cheaper model from the same provider only when that model suits the task. When in doubt, it keeps the model you asked for. Switch each strategy on or off; your own rules run first.
You choose; Auto Router add-onYour app calls a model name; the Auto Router add-on can select from models you configure.5
Prompt storage
Not storedMakas does not store your prompts or responses. Request logs keep usage metadata such as the model, tokens and cost, never the text. How it works
In your gateway; off unless turned onOff by default; can be turned on to keep prompts in your gateway's spend logs.6
Hosting
Hosted in the EUHosted by Makas in an EU region.
You run itSelf-hosted: you run it in your own environment.1
Also in Makas
More than a cheaper route.
Change the base URL
Add your Makas key. Nothing else. Keep the OpenAI, Anthropic or Google SDK; buffered and streaming requests are both routed.
Advanced Savings Analytics
Original and routed cost side by side, savings by model and by strategy, request logs and cost trends.
Routing management
Choose your routing strategies, add custom rules and set routing per key, with spending limits and budget alerts. All on the Founding Customer plan.
Sources
Every LiteLLM statement on this page comes from LiteLLM's own pages, listed here with the date each was checked.