Compare

Kunavo vs LiteLLM

LiteLLM is an open-source library and proxy server that gives you one OpenAI-compatible interface over 100+ providers — you run it and plug in your own Anthropic, Google and OpenAI keys, adding virtual keys, budgets, fallbacks and logging on top. Kunavo solves the other half: it is a hosted gateway that resells model access on one prepaid wallet below providers' list prices (varies by model), with nothing to deploy and no provider consoles to manage.

TL;DR

  • Both are OpenAI-compatible; the split is self-hosted BYO-key vs hosted resold inference.
  • LiteLLM: you run the SDK or proxy and bring your own provider keys; it unifies 100+ providers with budgets, virtual keys, fallbacks and logging. Open-source and self-hostable.
  • Kunavo: a hosted gateway — one prepaid wallet resells Claude, Gemini, GPT and image, video and audio models under list, nothing to deploy, one balance.
  • LiteLLM leads on provider breadth (incl. niche and self-hosted models), open-source control and no inference lock-in; Kunavo leads on price, zero-ops hosting, multimodal and the native Anthropic Messages API.
  • They compose: add Kunavo as an OpenAI-compatible provider in LiteLLM to keep its routing and budgets while inference runs through Kunavo's discounted catalog.
Side-by-side

Kunavo or LiteLLM?

CapabilityKunavoLiteLLM
OpenAI SDK drop-in
YesYes
Bring your own provider keys
LiteLLM routes through your provider accounts; Kunavo resells, so none are needed.
NoYes
Resells inference (one wallet)
YesNo
Price vs provider list
Most of the Kunavo catalog sits under list. With LiteLLM you pay each provider's list price.
−30% (uniform)Provider list
Nothing to deploy or run
Kunavo is a hosted API; LiteLLM's proxy is a service you host and operate.
YesNo
Open-source, self-hostable
NoYes
Provider breadth
LiteLLM reaches 100+ incl. niche and self-hosted; Kunavo curates a smaller frontier + multimodal shelf — /models is the live list.
Curated frontier100+ (incl. niche)
Virtual keys, budgets & spend caps
YesYes
Image / video / audio generation
First-class generation endpoints on Kunavo; LiteLLM mainly unifies text/embeddings.
YesPartial
Native Anthropic Messages API
YesPartial
Billing model
Kunavo: prepaid wallet from $10. LiteLLM: self-host free + you pay providers (or a Cloud plan).
Prepaid wallet from $10Self-host + provider bills
Local payment methods (Apple Pay / Pix / Alipay / WeChat)
Kunavo: one Stripe checkout. With LiteLLM the money goes to each provider you configure, separately.
YesEach provider separately

Where LiteLLM fits

LiteLLM is the strongest choice when you want an open-source unifying layer you fully control. If you already hold provider contracts, need to reach niche or self-hosted models, want to run the gateway inside your own VPC for compliance, and value a Python-native SDK plus a language-agnostic proxy with budgets and virtual keys, LiteLLM is built for exactly that. There is no inference lock-in — it simply standardises how your code calls whatever providers you configure.

Where Kunavo wins

Kunavo removes both the provider accounts and the ops. There is nothing to deploy: top up one wallet and call Claude, Gemini, GPT, plus image, video and audio, below each provider's list price depending on the model — no Anthropic Console, no Google Cloud project, no proxy to host and monitor, one balance. Generation endpoints are first-class, the native Anthropic Messages API is on the same key, and checkout is Stripe-native with local methods. Where LiteLLM gives you a gateway to run, Kunavo gives you a gateway to call.

How to pick (or use both)

Pick LiteLLM if you want an open-source, self-hosted gateway over provider keys you already manage, or need niche/self-hosted model coverage inside your own infrastructure. Pick Kunavo if you want zero-ops hosted inference on one prepaid wallet below list, first-class multimodal, and one balance to reconcile. They compose: register Kunavo as an OpenAI-compatible provider in LiteLLM (base URL https://api.kunavo.com/v1) to keep LiteLLM's routing and budgets while your tokens run through Kunavo's discounted catalog.

How you pay: LiteLLM vs Kunavo

LiteLLM is software rather than a merchant, so the payment question does not disappear — it moves one layer down. Kunavo's checkout is Stripe-native, so the list follows the buyer rather than our billing country: cards (Visa, Mastercard, Amex, JCB, UnionPay) and Apple Pay everywhere Stripe reaches, Link and Cash App Pay for one-tap repeat top-ups, Klarna and Amazon Pay, and Alipay and WeChat Pay for mainland China. Local rails appear by country — Pix in Brazil, Bancontact in Belgium, BLIK in Poland, EPS in Austria, MB WAY in Portugal — with Stripe converting the dollar amount into local currency on the checkout page, so a Brazilian buyer scans a Pix QR code for an amount in reais instead of hunting for a card cleared for international purchases. There is nothing to pay LiteLLM for beyond the servers you run it on; the money goes to each provider you configure. “Can I pay with Pix” therefore becomes “can I pay Anthropic with Pix”, and then the same question again for OpenAI and for Google — three accounts, three consoles, three card requirements. Kunavo collapses that into a single checkout, which is worth the most in exactly the places where an internationally-cleared card is hardest to get.

FAQ

Is LiteLLM an AI gateway?

Yes. LiteLLM is an open-source LLM gateway available as a Python SDK and a self-hosted proxy server, giving one OpenAI-compatible interface over 100+ providers. You run it and use your own provider keys rather than reselling inference.

Does LiteLLM host models or do I pay providers?

LiteLLM does not host or resell models — you configure your own provider keys and pay each provider at list price; the software is open-source and free to self-host. Kunavo is the opposite: it resells model access on one prepaid wallet below list (varies by model), with nothing to run.

Can I use LiteLLM and Kunavo together?

Yes. Add Kunavo as an OpenAI-compatible provider in LiteLLM (base URL https://api.kunavo.com/v1). You keep LiteLLM's routing, budgets and virtual keys while inference runs through Kunavo's discounted, multimodal catalog on one wallet.

Five minutes to switch one base_url.

If you're already on LiteLLM, switching to Kunavo is one base_url change — up to 70% below official rates on most models, pay as you go.