# A LiteLLM Alternative You Do Not Host

LiteLLM is open-source software you deploy on your own server — Docker, config files, provider keys, uptime your problem. RunAPI is a hosted gateway: same OpenAI-compatible API, same 100+ LLMs, no server to run. Change the base URL and your existing code keeps working. RunAPI also adds image, video, and music generation, built-in billing, MCP server, CLI, and SDKs at 50% off official rates.

*Updated June 23, 2026 · RunAPI Editorial*

## LiteLLM vs RunAPI — self-hosted or managed?

LiteLLM is an open-source Python proxy you run on your own server. It supports 100+ LLMs through a unified OpenAI-compatible interface — you pay provider costs and your own hosting, but you control everything. RunAPI is the managed version of that idea: the same OpenAI-compatible API, no server to run, 50% off official model rates, and coverage extended to image, video, and music generation under one API key. If the setup and ops work of self-hosting isn't something you want to own, RunAPI gets you the same multi-model access without it.

- **Hosting model**: LiteLLM runs on your infra (Docker, VMs, K8s). RunAPI is a hosted service — send requests to runapi.ai, no deployment required.
- **Modality coverage**: LiteLLM routes text LLMs. RunAPI routes LLMs and adds image, video, music, and audio generation under one key.
- **Pricing**: LiteLLM is free software; you pay provider API costs plus hosting. RunAPI charges 50% of official model rates with no infra overhead.
- **Developer toolkit**: RunAPI ships an MCP server, CLI, and SDKs for Python, Node.js, PHP, Java, Ruby, and Go alongside full OpenAI-SDK compatibility.

## LiteLLM vs RunAPI feature breakdown

The table compares the two gateways across the dimensions that matter most: whether you want to run your own server or just use an API, how much you pay, and what models you can actually reach.

| Feature | LiteLLM | RunAPI |
|---|---|---|
| Hosting model | Self-hosted (your infra) | Fully managed service |
| LLM routing | Yes — 100+ models | Yes — 100+ models |
| Image generation | No | Yes |
| Video generation | No | Yes |
| Music and audio | No | Yes |
| OpenAI-compatible API | Yes | Yes |
| Built-in billing | Manual (BYO keys) | Yes — unified account billing |
| MCP server | No | Yes |
| Official SDKs | Python (LiteLLM SDK) | Python, Node.js, PHP, Java, Ruby, Go + OpenAI-compatible |
| CLI | litellm CLI (server mgmt) | Yes — generation and key management |
| Cost | Provider cost + hosting | 50% off official provider rates |

## What changes when you drop the self-hosted proxy?

Running LiteLLM means cloning the repo, editing a .env file with a master key and salt key, running docker-compose, then configuring each provider through the admin UI — and that's just the first time. After that you own the uptime, the upgrades, and the key rotation. RunAPI removes that layer entirely. Send requests to runapi.ai and pay per call. No server, no config files, no provider accounts to juggle.

### No server to provision

LiteLLM requires a running Python process or Docker container on a VPS you pay for and monitor. RunAPI is an HTTPS endpoint — point your client at it and start calling.

### No config files to maintain

LiteLLM's config.yaml maps model names to provider keys and handles fallbacks. RunAPI handles routing internally; you supply one API key.

### No key management per provider

With LiteLLM you store Anthropic, OpenAI, Gemini, and other keys on your server. RunAPI holds one key on your end — it manages provider credentials on its side.

### Unified billing in one place

LiteLLM gives you raw provider invoices across multiple accounts. RunAPI consolidates all usage into a single dashboard with per-call cost records — one bill instead of a bunch of paid API accounts.


## Which models does RunAPI route?

LiteLLM's strength is breadth — it routes 100+ LLMs from every major provider. RunAPI matches that LLM coverage across the frontier models most teams actually use (Claude, GPT, Gemini, DeepSeek) and adds generative media models not available through LiteLLM.

### Language models

Claude, GPT, Gemini, DeepSeek, and other LLMs for chat, coding, reasoning, and tool use — the same text coverage LiteLLM users expect from the frontier models.

### Image models

Text-to-image and image-edit models including Flux and Seedream, callable from the same API key as your LLM calls.

### Video models

Text-to-video and image-to-video models for generating short clips, billed pay-as-you-go per task.

### Music and audio

Music generation and audio models for soundtracks and voice work — modalities outside LiteLLM's scope entirely.


## How does RunAPI pricing compare to self-hosting LiteLLM?

LiteLLM is free software but you pay full provider rates on top of what your VPS costs (~$6–12/month) plus time spent maintaining the setup. RunAPI charges 50% of official provider rates with no hosting overhead. For teams that aren't optimizing LiteLLM's advanced routing features, the total cost often favors the managed option.

| Model | Official input /M | Official output /M | RunAPI input /M | RunAPI output /M |
|---|---|---|---|---|
| Claude Sonnet 4.6 | $6.00 | $30.00 | $3.00 | $15.00 |
| Claude Opus 4.7 | $10.00 | $50.00 | $5.00 | $25.00 |
| GPT-5.4 | $2.50 | $15.00 | $1.25 | $7.50 |
| Gemini 2.5 Pro | $1.25 | $10.00 | $0.63 | $5.00 |

## How to switch from LiteLLM to RunAPI

1. **Create a RunAPI account** — Sign up at runapi.ai. No credit card required to get started.
2. **Get your API key** — Go to Dashboard -> API Keys, create a key, and copy it to your environment.
3. **Update the base URL** — In your OpenAI-compatible client or LiteLLM config, change the base URL to https://runapi.ai/v1 and replace your LiteLLM key with the RunAPI key. No other code changes are needed.
4. **Decommission your LiteLLM server** — Once requests are routing through RunAPI, you can stop your LiteLLM process and remove the server. Provider keys stored there are no longer needed.

## LiteLLM Alternative FAQ

### Isn't this still paying for cloud APIs? How is this different from just using the services directly?

Good question — the difference is one API key, one bill, and one base URL. With LiteLLM you still need an Anthropic key, an OpenAI key, a Gemini key, and so on, plus the LiteLLM server to hold them all together. RunAPI gives you a single key that reaches every model. You drop it in where you used to put your OpenAI key, change the base URL to runapi.ai/v1, and you're done. No accounts to manage per provider, no invoices per service.

### You started with self-hosted, then immediately added a bunch of paid cloud API subscriptions — isn't that the opposite of self-hosted?

LiteLLM gives you self-hosting with full control — your server, your provider keys, your config files. RunAPI is the other side of that choice: no Docker, no VPS to pay for, no config.yaml to maintain, no uptime to monitor. You still get the same OpenAI-compatible API and access to 100+ LLMs. The tradeoff is that you don't run the infrastructure; RunAPI does. If control and privacy are the priority, LiteLLM makes sense. If you want multi-model access without the setup, RunAPI gets you there faster.

### Is RunAPI a drop-in replacement for LiteLLM?

RunAPI is a drop-in replacement for the OpenAI-compatible API surface — the same /v1/chat/completions endpoint. Change the base URL and API key in your existing client and requests start routing through RunAPI. LiteLLM-specific features like virtual keys and custom routing rules have no equivalent, so check those before switching.

### How much does RunAPI cost compared to LiteLLM?

LiteLLM is free software but you pay full provider rates plus the cost of running a server (a VPS typically runs $6–12/month) plus time spent on upgrades, config changes, and debugging outages. RunAPI charges 50% of official model rates with no hosting overhead. For Claude Sonnet 4.6, that is $3 per million input tokens and $15 per million output tokens versus $6 and $30 at official rates — and nothing extra for the server.

### How long does it take to migrate from LiteLLM to RunAPI?

Most teams finish in under 15 minutes. The change is one line — replace your LiteLLM proxy URL with https://runapi.ai/v1 and swap your LiteLLM key for a RunAPI key. If you use the OpenAI Python SDK, no other code changes are needed. After confirming requests route through RunAPI, shut down the LiteLLM Docker container. The same RunAPI key also works with Claude Code, Cursor, Windsurf, and the MCP server.

### What is the MCP server and how does it differ from LiteLLM?

RunAPI ships a Model Context Protocol server that lets MCP-aware hosts like Claude Code discover models, check pricing, and create tasks directly from the assistant interface. LiteLLM has no MCP server. The RunAPI MCP server is additive — you can use it alongside the REST API rather than instead of it.

### Can I use the OpenAI Python SDK with RunAPI?

Yes. Point the base_url parameter at https://runapi.ai/v1 and pass your RunAPI key as api_key. The rest of your OpenAI SDK code stays unchanged. RunAPI also publishes its own typed SDKs for Python, Node.js, PHP, Java, Ruby, and Go if you prefer a client built for RunAPI's full API surface including media generation.

### What happens to LiteLLM features like virtual keys and spend tracking?

LiteLLM's virtual key system lets you issue internal keys with per-key spend caps and model restrictions — useful if you're running LiteLLM as a shared hub for a family or team. RunAPI doesn't have virtual keys, but provides per-account usage reporting and per-call cost records in the dashboard. If the main reason you use LiteLLM's virtual keys is to track or limit spend, RunAPI's account billing covers that. If you need to issue keys to other people with different model access, evaluate that gap before migrating.