Agent Research

Google Deep Research models run long, research-focused agent jobs via /v1/research/runs.

Example

curl https://llmtr.com/v1/research/runs \
  -H "Authorization: Bearer sk_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/deep-research-pro-preview-12-2025",
    "input": "Compare the latest frontier reasoning models for coding agents.",
    "max_budget_usd": 2.00
  }'

Response

{
  "id": "run_xxx",
  "status": "succeeded",
  "output": {
    "summary": "...",
    "sources": [
      { "url": "https://...", "title": "..." }
    ]
  },
  "usage": {
    "prompt_tokens": 12400,
    "completion_tokens": 4200,
    "total_tokens": 16600
  }
}

Parameters

Field	Description
input	Research question or brief
max_budget_usd	Spend cap
context	Optional extra context (string or file_id array)

Long-running support

Runs may return pending. Poll with:

curl https://llmtr.com/v1/research/runs/run_xxx \
  -H "Authorization: Bearer sk_your_key"