← Back to Docs
kr-compare
Run the same prompt across multiple models in parallel. Compare outputs, latency, cost, and quality scores side by side. Perfect for evaluations, model selection, and benchmarking.
Paid accounts only, you're billed for each model's actual token usage.
Endpoint
POST /api/v1/compare
Request Body
{
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Explain quantum computing in one paragraph."}
],
"models": [
"claude-sonnet-4-6",
"gpt-5.4",
"gemini-3.1-pro-preview"
],
"temperature": 0.7,
"max_tokens": 512
}Parameters
messages. Standard OpenAI-format messages array. Required.
models. Array of model IDs to compare (max 5). Optional, defaults to one model from each quality tier (premium, mid, budget).
temperature. Sampling temperature. Optional, default 0.7.
max_tokens. Max output tokens per model. Optional, default 1024.
Response
{
"object": "comparison",
"task_classification": {
"category": "reasoning",
"confidence": 0.87
},
"models_compared": 3,
"total_cost_usd": 0.0234,
"comparisons": [
{
"status": "success",
"model": "claude-sonnet-4-6",
"provider": "Anthropic",
"displayName": "Claude Sonnet 4.6",
"qualityScore": 0.96,
"content": "Quantum computing harnesses...",
"usage": {
"prompt_tokens": 24,
"completion_tokens": 156,
"total_tokens": 180
},
"costUsd": 0.0081,
"latencyMs": 1240,
"finishReason": "stop"
},
...
]
}Each model runs in parallel so total latency is roughly equal to the slowest model, not the sum. Failed models appear with status: "error" and an error message.
cURL Example
curl https://api.kairosroute.com/v1/compare \
-H "Authorization: Bearer kr-your-key" \
-H "Content-Type: application/json" \
-d '{
"messages": [{"role": "user", "content": "Write a haiku about APIs"}],
"models": ["claude-sonnet-4-6", "gpt-5.4-mini", "deepseek-chat"]
}'