AI Model Cost Per Task Benchmark (Multiple Models)

{
  "type": "object",
  "required": [
    "models"
  ],
  "properties": {
    "models": {
      "type": "string",
      "description": "Required. Comma-separated list of model slugs to fetch (e.g. 'claude-sonnet-5,gpt-5-6-sol,gemini-3-5-flash-lite')."
    }
  },
  "additionalProperties": false
}

{
  "type": "object",
  "properties": {
    "data": {
      "type": "object",
      "properties": {
        "models": {
          "type": "array",
          "items": {
            "type": "object",
            "properties": {
              "name": {
                "type": "string",
                "description": "Full display name, e.g. 'GPT-5.6 Sol (max)'."
              },
              "slug": {
                "type": "string",
                "description": "The model slug."
              },
              "ttft": {
                "type": "number",
                "description": "Time to first token, seconds; populated on every model observed across both real calls."
              },
              "short_name": {
                "type": "string",
                "description": "Short display name; often identical to `name`, sometimes a shortened form."
              },
              "coding_index": {
                "type": [
                  "number",
                  "null"
                ],
                "description": "Coding-specific benchmark score; null on every model observed on both real calls -- the benchmark does not currently expose per-coding-task data."
              },
              "cost_per_task": {
                "type": "number",
                "description": "USD, average total cost to run the benchmark's standard task."
              },
              "time_per_task": {
                "type": "number",
                "description": "Seconds, average total time to complete the standard task; populated on every model observed across both real calls (7 distinct models total)."
              },
              "intelligence_index": {
                "type": "number",
                "description": "The benchmark's aggregate reasoning/quality score for this model."
              },
              "coding_cost_per_task": {
                "type": [
                  "number",
                  "null"
                ],
                "description": "USD, coding-task variant of `cost_per_task`; null on every model observed."
              },
              "coding_time_per_task": {
                "type": [
                  "number",
                  "null"
                ],
                "description": "Seconds, coding-task variant of `time_per_task`; null on every model observed."
              },
              "input_tokens_per_task": {
                "type": "number"
              },
              "output_tokens_per_task": {
                "type": "number"
              },
              "cost_per_task_breakdown": {
                "type": "object",
                "properties": {
                  "cache_hit": {
                    "type": "number"
                  },
                  "cache_write": {
                    "type": "number"
                  },
                  "input_tokens": {
                    "type": "number"
                  },
                  "output_tokens": {
                    "type": "number"
                  },
                  "reasoning_tokens": {
                    "type": "number"
                  }
                },
                "description": "USD contribution of each token category to `cost_per_task`.",
                "additionalProperties": true
              },
              "end_to_end_response_time": {
                "type": "number",
                "description": "Seconds; populated on every model observed across both real calls."
              },
              "reasoning_tokens_per_task": {
                "type": "number"
              },
              "coding_cost_per_task_breakdown": {
                "type": [
                  "object",
                  "null"
                ],
                "description": "Coding-task variant of `cost_per_task_breakdown`; null on every model observed. Populated shape unconfirmed since no non-null example exists.",
                "additionalProperties": true
              }
            },
            "additionalProperties": true
          },
          "description": "One entry per requested model."
        }
      },
      "additionalProperties": true
    },
    "status": {
      "type": "string",
      "description": "'success' on both real calls observed."
    }
  },
  "description": "The cost-per-task payload, as returned at routing.responseResultPath ($.output).",
  "additionalProperties": true
}