Sign in to start generating with your available language models.
Call Claude/claude-fable-5 through /v1/messages.
Keep the same RunAPI key and request shape while comparing outputs.
Call Claude/claude-fable-5-1 through /v1/messages.
Keep the same RunAPI key and request shape while comparing outputs.
Call Claude/claude-haiku-4-5-20251001 through /v1/messages.
Keep the same RunAPI key and request shape while comparing outputs.
The OpenAI shape every SDK already speaks. Swap base_url and keep your existing request code.
The Messages API shape for system prompts, tool use, and structured content blocks.
The stateful shape for multi-turn agent loops with server-side conversation handling.
Vector embeddings for retrieval, billed per token on the same RunAPI key.
Anthropic
$10.00 / 1M in
Anthropic
$10.00 / 1M in
Anthropic
$0.60 / 1M in
Anthropic
$9.00 / 1M in
Anthropic
$3.00 / 1M in
Anthropic
$3.00 / 1M in
Anthropic
$3.00 / 1M in
Anthropic
$3.00 / 1M in
curl https://runapi.ai/v1/chat/completions \
-H "Authorization: Bearer $RUNAPI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"Explain the difference between concurrency and parallelism."}]}'Create one key for your RunAPI account.
Send requests through the documented OpenAI-compatible endpoint.
RunAPI selects a supported route for your request.
Receive streamed output when the selected endpoint supports SSE.
Stop juggling API keys. One integration gives you access to 230+ models across all providers.
Our volume agreements with providers mean lower prices than going direct. No markup on top.
If a provider goes down, we route to the next best option. Your app stays up.
Drop-in replacement for OpenAI SDK. Change one line of code to access any model.
Pay only for what you use. Per-request pricing with no minimums. Track costs per model, per project.
Our Privacy Policy explains how request data and generated content are processed and retained.
Route planning to Opus and edits to a cheap model, on one key with one invoice.
Embeddings and generation on the same endpoint, so retrieval and answering share a budget.
Fan out drafts across models and keep the best, without signing five provider contracts.
A/B the same prompt across Claude, GPT, and Gemini by changing one string.
| Feature | RunAPI | OpenRouter | Direct API |
|---|---|---|---|
| LLM models | 66 | 300+ | 1 provider |
| Other modalities | Video, Image, Music, Audio | No | Per provider |
| Pricing | 15-25% below official | Market rate | Official rate |
| Native Messages API | Yes | No | Yes |
| Unified credits | Yes | Yes | No |
| MCP server + CLI | Yes | No | No |
Point the client at the documented RunAPI base URL and use your RunAPI key. Keep the request shape supported by the selected endpoint.
Streaming is available when the selected model and endpoint expose a streaming response.
Tool calling is available on models and endpoint shapes that declare tool support in their runtime contract.
Input and output rates are listed in the model catalog. Usage is billed from the token counts reported for the request.
Choose the endpoint shape that matches your existing client and the capabilities required by your request.
The API returns the request status through its documented response contract so your application can retry or choose another supported model.
67 models behind the OpenAI shape you already ship. Start free, no credit card required.