{
  "schema": "anvil-serving.controlled-ab-workloads/v1",
  "preflight": {
    "checks": [
      "smoke",
      "json",
      "needle",
      "tools",
      "long-tools",
      "streaming-tools",
      "tool-result",
      "responses",
      "image",
      "ocr"
    ],
    "max_completion_tokens": 65536,
    "reasoning_effort": "max",
    "ocr_note": "Baseline combined OCR fixture was invalid; retain it and compare the corrected dedicated OCR observation."
  },
  "vision": {
    "observations": 18,
    "unique_fixture_images": 8,
    "max_images_per_request": 8,
    "concurrency": 1,
    "max_completion_tokens": 9216,
    "temperature": 1.0,
    "top_p": 0.95
  },
  "agentic": {
    "case_ids": [
      "planning",
      "reasoning",
      "structured-edit",
      "tool-sequence",
      "parallel-tools",
      "dependent-result",
      "tool-recovery",
      "debug-loop",
      "context-recovery"
    ],
    "oracle_revision": "8",
    "instrument_sha256": "1f40041cbf13275ef3635f514fb8b23cbf23504778f79c1a02b0aeb2211132f6",
    "profile_sha256": "fce0e6e1e7497585bdfc6f76507d3baaff99a6b4f174d841d268de20c68de7f6",
    "max_steps": 24,
    "max_completion_tokens": 8192,
    "thinking_mode": "default",
    "reasoning_effort": "max",
    "temperature": 0.0,
    "top_p": 0.95,
    "repetitions": 2,
    "campaign_required_pass_rate": 1.0
  },
  "context": {
    "case_ids": [
      "native-needle",
      "native-order",
      "native-distractor"
    ],
    "token_buckets": [
      32768,
      318464
    ],
    "positions": [
      0.1,
      0.5,
      0.9
    ],
    "repetitions": 1,
    "output_headroom_tokens": 9216,
    "advertised_context": 327680,
    "thinking_mode": "default",
    "reasoning_effort": "max",
    "temperature": 1.0,
    "top_p": 0.95
  },
  "stability": {
    "scenario": "baseline-stability-scout-r3.json",
    "anchor_tokens": 32000,
    "contender_tokens": 8192,
    "anchor_max_tokens": 16384,
    "contender_max_tokens": 8192,
    "reasoning_effort": "max",
    "temperature": 0.0,
    "top_p": 0.95,
    "rounds": 1,
    "purpose": "bounded overlap/retrieval diagnostic; not scheduler proof, endurance, or performance ranking"
  },
  "timing": {
    "cells": [
      {
        "context_tokens": 8192,
        "concurrency": 1,
        "seed": 941111
      },
      {
        "context_tokens": 8192,
        "concurrency": 4,
        "seed": 941112
      },
      {
        "context_tokens": 32768,
        "concurrency": 1,
        "seed": 941113
      },
      {
        "context_tokens": 32768,
        "concurrency": 4,
        "seed": 941114
      }
    ],
    "requests_per_cell": 8,
    "max_completion_tokens": 16384,
    "prompt_cache_mode": "unique",
    "response_words": 0,
    "controlled_output_policy": "observe",
    "temperature": 0.6,
    "top_p": 0.95,
    "reasoning_effort": "max",
    "scope": "Natural completion latency and observed generation; output lengths may differ. Eight observations per cell do not establish tail-latency SLOs."
  }
}