starter.wikiSearch⌘K

Self-operated OpenAI-compatible AI proxy

LiteLLM

An open-source proxy and control plane that presents OpenAI-compatible request patterns across many model providers with routing, budgets, keys, and observability integrations.

Visit official site

Editorial verdict

Choose LiteLLM when a self-operated proxy boundary is a requirement and the platform team can own reliability, credentials, upgrades, telemetry, policy, and provider-specific compatibility.12

Best for

  • Self-operated multi-provider proxying
  • Platform-owned keys, budgets, and routing
  • Teams willing to test provider-specific behavior
12

Not ideal for

  • Teams without a critical-proxy operator
  • Products needing a fully managed control plane
  • Buyers assuming one compatibility layer normalizes every provider feature
12
Main trade-off

Open-source proxy control reduces application coupling while transferring availability, security, upgrades, logging, compatibility, and provider-change response to the platform team.12

Product boundary

Whether self-operating LiteLLM's proxy and control plane is justified by provider routing and governance needs despite compatibility and infrastructure ownership.

This page evaluates the LiteLLM proxy and control plane, not every use of its Python client. OpenAI-compatible request shapes do not make provider models, parameters, streaming, tool use, errors, safety, or policies interchangeable.12

For: Platform teams that need a self-operated multi-provider proxy and can own availability, credentials, upgrades, logs, and compatibility testing

  • The production workload, language, model, provider, or deployment boundary changes
  • Current pricing, retention, data use, policy, or regional support changes
  • Measured quality, latency, reliability, or operating cost no longer fits

Why teams consider LiteLLM

  • Compatible entry pointThe proxy exposes common OpenAI-style request patterns for many providers.12
  • Control-plane primitivesVirtual keys, budgets, rate limits, routing, fallbacks, and logging are documented.12
  • Self-operated routeThe OSS proxy lets teams own the gateway data and runtime boundary.12

Pricing

The open-source proxy has no represented per-token software fee; infrastructure and operations remain customer costs, while Enterprise terms are based on contracted gateway capacity and support.1

Current official route

Open-source proxy or Enterprise

Verified 2026-07-27: LiteLLM publishes an OSS self-hosted route and custom Enterprise pricing based on gateway request capacity, deployment, and support rather than representing a universal token markup.1

OSS cost
Compute, database, cache, networking, observability, security, upgrades, and staffing1
Enterprise boundary
Contracted request capacity, deployment, support, and features1
Provider cost
Underlying model usage remains separately billed1
Pricing checked View official pricing

LiteLLM vs alternatives

Portkey

Choose when
Choose Portkey when a managed AI control plane removes meaningful routing and governance work, after modeling provider bills, gateway traffic, recorded logs, retention, privacy, and deployment.
Avoid when
Teams requiring a self-operated gateway by default
Compared with LiteLLM
A managed control plane centralizes reliability and governance while adding request, log, retention, data, deployment, and platform dependence on top of provider inference.34

OpenRouter

Choose when
Choose OpenRouter when hosted model and provider choice is more valuable than direct-provider control, after governing fees, credits or BYOK, routing, provider logging, retention, and fallback behavior.
Avoid when
Teams requiring self-hosted gateway control
Compared with LiteLLM
Broad hosted model access reduces integration work while adding aggregator fees, credits, provider-routing, privacy, availability, and another commercial dependency.56

Vercel AI Gateway

Choose when
Choose Vercel AI Gateway when its managed endpoint and provider operations reduce meaningful work, after verifying current pricing, credits, BYOK, failover, provider order, and ZDR eligibility.
Avoid when
Teams requiring a self-operated gateway
Compared with LiteLLM
A managed Vercel gateway simplifies provider access and failover while adding gateway, credit, catalog, routing, data-policy, and platform dependence.78

Resources and sources

Official product, documentation, pricing, and policy sources

  • LiteLLM pricing
    Open
  • LiteLLM proxy documentation
    Open
  • Portkey pricing
    Open
  • Portkey AI Gateway documentation
    Open
  • OpenRouter pricing and billing FAQ
    Open
  • OpenRouter provider routing
    Open
  • Vercel AI Gateway product and pricing
    Open
  • Vercel AI Gateway documentation
    Open
  1. 1
    LiteLLM pricing

    LiteLLM · Accessed Official

  2. 2
    LiteLLM proxy documentation

    LiteLLM · Accessed Official

  3. 3
    Portkey pricing

    Portkey · Accessed Official

  4. 4
    Portkey AI Gateway documentation

    Portkey · Accessed Official

  5. 5
    OpenRouter pricing and billing FAQ

    OpenRouter · Accessed Official

  6. 6
    OpenRouter provider routing

    OpenRouter · Accessed Official

  7. 7
    Vercel AI Gateway product and pricing

    Vercel · Accessed Official

  8. 8
    Vercel AI Gateway documentation

    Vercel · Accessed Official