Documentation Index
Fetch the complete documentation index at: https://docs.routing.run/llms.txt
Use this file to discover all available pages before exploring further.
POST /v1/chat/completions matches OpenAI Chat Completions. Point the SDK base_url at routing.run.
This is the recommended default endpoint for apps, SDKs, and coding agents.
Base URL
Primary:
https://api.routing.run/v1
https://api.routing.run is slow or returning errors. It is hosted by the routing.run team and supports the same API keys and paths.
Quick setup
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ROUTING_RUN_API_KEY"],
base_url="https://api.routing.run/v1",
)
response = client.chat.completions.create(
model="route/deepseek-v3.2",
messages=[
{"role": "system", "content": "You are a staff engineer reviewing a pull request."},
{"role": "user", "content": "List concrete issues in this diff: …"},
],
temperature=0.2,
)
print(response.choices[0].message.content)
Request
curl -sS -X POST https://api.routing.run/v1/chat/completions \
-H "X-API-Key: ${ROUTING_RUN_API_KEY}" \
-H "Content-Type: application/json" \
-d '{
"model": "route/minimax-m2.7",
"messages": [
{"role": "user", "content": "Hello"}
]
}'
Chat models
| Route | Provider | Context |
|---|
route/minimax-m2.5 | minimax/opencode | 100k |
route/minimax-m2.7 | opencode/minimax | 100k |
route/kimi-k2.5 | crof | 131k |
route/kimi-k2.6 | crof | 131k |
route/glm-5 | crof | 200k |
route/deepseek-v3.2 | crof/chutes | 163k |
route/deepseek-v4-pro | crof | 163k |
route/qwen3.5-plus | opencode | 100k |
route/mistral-large-3 | routing-inference | See model metadata |
route/mistral-medium-2505 | routing-inference | See model metadata |
route/mistral-small-2503 | routing-inference | See model metadata |
Mistral chat and summarization
Use Mistral models through the same POST /v1/chat/completions endpoint.
Chat request
{
"model": "route/mistral-large-3",
"messages": [
{
"role": "user",
"content": "Write a concise product announcement for routing.run embeddings."
}
]
}
Summarization request
{
"model": "route/mistral-medium-2505",
"messages": [
{
"role": "system",
"content": "Summarize the user's text into 5 bullet points."
},
{
"role": "user",
"content": "<long document text here>"
}
],
"max_tokens": 700
}
Mistral models
| Model | Use case |
|---|
route/mistral-large-3 | Premium chat completion and writing |
route/mistral-medium-2505 | Summarization and balanced generation |
route/mistral-small-2503 | Fast summarization and lightweight generation |
TypeScript example
async function summarizeWithMistral(text: string) {
const response = await fetch(`${API_BASE_URL}/v1/chat/completions`, {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'X-API-Key': ROUTING_API_KEY,
},
body: JSON.stringify({
model: 'route/mistral-medium-2505',
messages: [
{ role: 'system', content: 'Summarize the user text into 5 bullet points.' },
{ role: 'user', content: text },
],
max_tokens: 700,
}),
})
if (!response.ok) {
throw new Error('Mistral summarization request failed')
}
return response.json()
}
Request body
Response
{
"id": "chatcmpl_01J8rQvN4pK2mL9xYz3wAbcDef",
"object": "chat.completion",
"created": 1744701234,
"model": "route/deepseek-v3.2",
"latency_ms": 842,
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": "- Line 42: possible null dereference …",
"reasoning_content": "Internal reasoning trace may appear here on some models"
},
"finish_reason": "stop"
}
],
"usage": {
"prompt_tokens": 128,
"completion_tokens": 256,
"total_tokens": 384
}
}
latency_ms, provider, and on some models message.reasoning_content are routing.run extensions on top of the OpenAI completion object.
Live checks against route/glm-5.1-precision, route/qwen3.6-plus, and route/kimi-k2.5 showed that some reasoning-oriented models may spend max_tokens on reasoning first. If you set a very small max_tokens, message.content can be empty while message.reasoning_content is populated and finish_reason may be length.
Streaming
Set "stream": true to receive a streaming response:
stream = client.chat.completions.create(
model="route/deepseek-v3.2",
messages=[{"role": "user", "content": "Stream a short design for a rate limiter."}],
stream=True,
)
for chunk in stream:
if chunk.choices[0].delta.content:
print(chunk.choices[0].delta.content, end="")
tools = [
{
"type": "function",
"function": {
"name": "get_weather",
"description": "Get current weather for a location",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "City name"
}
},
"required": ["location"]
}
}
}
]
response = client.chat.completions.create(
model="route/deepseek-v3.2",
messages=[{"role": "user", "content": "What's the weather in Tokyo?"}],
tools=tools,
tool_choice="auto",
)
Error handling
Non-streaming failures from routing logic use plain text bodies and X-Error-Code (see Authentication). Streaming failures may emit an SSE data: line with JSON {"error":{"message":"…","type":"api_error"}}.
| Status | X-Error-Code (typical) | Meaning |
|---|
| 400 | INVALID_MODEL | Unknown route/… id |
| 401 | AUTHENTICATION_ERROR | Bad or missing rk_ / JWT on inference routes |
| 403 | MODEL_NOT_ALLOWED | Plan cannot call this model |
| 429 | DAILY_REQUEST_LIMIT_EXCEEDED | Daily request cap |
| 502 | PROVIDER_ERROR | Every upstream in the routing chain failed |
| 504 | PROVIDER_TIMEOUT | Upstream timeouts exhausted |
