{
  "schema_version": 1,
  "captured_at_utc": "2026-07-28",
  "model_requested": "glm-5.2",
  "model_returned": "glm-5.2",
  "route": {
    "product": "Z.ai Coding Plan",
    "endpoint_base": "https://api.z.ai/api/coding/paas/v4",
    "scope": "Dedicated coding-tool route; not tested or recommended as a general application backend."
  },
  "container": {
    "image": "ghcr.io/astral-sh/uv@sha256:41977070f8f7a569ba48da4641bf5212421625e02763bbca24d6fe599ac62a91",
    "user": "1000:1000",
    "root_filesystem": "read-only",
    "capabilities": "all dropped",
    "memory": "256m",
    "cpus": 1,
    "pids": 64,
    "network": "host networking for outbound HTTPS only; no listener or published port"
  },
  "method": {
    "measured_requests": 44,
    "excluded_warmups": 3,
    "sampling": "do_sample=false",
    "unique_nonce_at_prompt_start": true,
    "cached_prompt_tokens_observed": [
      0
    ],
    "p50_method": "median",
    "p90_method": "nearest-rank",
    "reasoning_effort_requested": "low",
    "reasoning_effort_documented_mapping": "Z.ai documents low and medium as mapped to high for GLM-5.2.",
    "sensitive_text_stored": false
  },
  "summary": {
    "strict_thinking_disabled_sequential": {
      "attempts": 10,
      "successes": 10,
      "first_content_ms": {
        "p50": 2972.23,
        "p90": 3794.952,
        "min": 2387.818,
        "max": 3900.619
      },
      "total_ms": {
        "p50": 3611.895,
        "p90": 4471.389,
        "min": 2856.594,
        "max": 4863.947
      },
      "client_stream_phase_tokens_per_second": {
        "definition": "API completion_tokens divided by time from first visible content to [DONE]. Includes delivery behavior and is not accelerator decode throughput.",
        "p50": 186.072,
        "min": 97.274,
        "max": 245.32
      },
      "required_marker_successes": 10
    },
    "strict_thinking_disabled_concurrency_3": {
      "attempts": 9,
      "successes": 9,
      "first_content_ms": {
        "p50": 4089.834,
        "p90": 6927.126,
        "min": 3080.951,
        "max": 6927.126
      },
      "total_ms": {
        "p50": 4608.762,
        "p90": 8019.609,
        "min": 3532.665,
        "max": 8019.609
      },
      "change_from_sequential_percent": {
        "first_content_p50": 37.6,
        "first_content_p90": 82.5,
        "total_p50": 27.6,
        "total_p90": 79.4
      },
      "required_marker_successes": 8
    },
    "strict_thinking_enabled_requested_low_max_384": {
      "attempts": 10,
      "http_sse_successes": 10,
      "finish_reason_length": 10,
      "visible_content_responses": 0,
      "completion_tokens_each": 384,
      "reasoning_token_range": [
        379,
        383
      ],
      "first_reasoning_ms": {
        "p50": 2772.702,
        "p90": 3682.101,
        "min": 1974.371,
        "max": 4814.432
      },
      "total_ms": {
        "p50": 6677.322,
        "p90": 8930.369,
        "min": 5032.544,
        "max": 8990.711
      }
    },
    "strict_thinking_enabled_requested_low_max_1024": {
      "attempts": 5,
      "http_sse_successes": 5,
      "finish_reason_length": 5,
      "visible_content_responses": 0,
      "completion_tokens_each": 1024,
      "reasoning_token_range": [
        1020,
        1024
      ],
      "first_reasoning_ms": {
        "p50": 2104.663,
        "p90": 2432.589,
        "min": 1854.482,
        "max": 2432.589
      },
      "total_ms": {
        "p50": 10065.797,
        "p90": 12439.008,
        "min": 9077.677,
        "max": 12439.008
      }
    },
    "relaxed_thinking_disabled": {
      "attempts": 5,
      "successes": 5,
      "first_content_ms": {
        "p50": 3461.158,
        "p90": 5038.592,
        "min": 2876.756,
        "max": 5038.592
      },
      "total_ms": {
        "p50": 3708.139,
        "p90": 5084.597,
        "min": 3249.976,
        "max": 5084.597
      },
      "required_marker_successes": 4
    },
    "relaxed_thinking_enabled_requested_low": {
      "attempts": 5,
      "successes": 5,
      "first_reasoning_ms": {
        "p50": 3474.033,
        "p90": 4501.577,
        "min": 2869.501,
        "max": 4501.577
      },
      "first_content_ms": {
        "p50": 3682.115,
        "p90": 5245.195,
        "min": 2869.715,
        "max": 5245.195
      },
      "reasoning_to_content_gap_ms": {
        "p50": 0.258,
        "p90": 1771.162,
        "min": 0.214,
        "max": 1771.162
      },
      "total_ms": {
        "p50": 5134.432,
        "p90": 5421.781,
        "min": 3693.005,
        "max": 5421.781
      },
      "required_marker_successes": 5
    }
  },
  "raw_timing_samples_ms": {
    "strict_sequential_first_content": [
      2964.757,
      3900.619,
      2549.123,
      2979.702,
      3794.952,
      2387.818,
      3784.517,
      2749.357,
      3695.137,
      2446.816
    ],
    "strict_sequential_total": [
      3765.896,
      4471.389,
      3067.906,
      3457.894,
      4423.147,
      2856.594,
      4863.947,
      3337.2,
      4442.198,
      3049.737
    ],
    "concurrency_3_first_content": [
      3080.951,
      3664.126,
      4635.652,
      4309.659,
      3154.751,
      4089.834,
      3920.453,
      4288.834,
      6927.126
    ],
    "concurrency_3_total": [
      3532.665,
      4307.168,
      5785.668,
      4875.899,
      4122.675,
      4135.273,
      4608.762,
      4814.123,
      8019.609
    ],
    "relaxed_thinking_enabled_first_reasoning": [
      3681.862,
      3327.765,
      4501.577,
      3474.033,
      2869.501
    ],
    "relaxed_thinking_enabled_first_content": [
      3682.115,
      3328.023,
      4502.21,
      5245.195,
      2869.715
    ]
  },
  "buffering_observations": [
    {
      "scenario": "concurrency-3",
      "sample": 6,
      "content_fragments": 102,
      "first_to_last_content_ms": 45.342,
      "naive_tokens_per_second": 2288.783
    },
    {
      "scenario": "relaxed-thinking-disabled",
      "sample": 2,
      "content_fragments": 95,
      "first_to_last_content_ms": 45.796,
      "naive_tokens_per_second": 2086.73
    },
    {
      "scenario": "relaxed-thinking-disabled",
      "sample": 3,
      "content_fragments": 80,
      "first_to_last_content_ms": 0.761,
      "naive_tokens_per_second": 89912.281
    }
  ],
  "limits": [
    "One account, one client host, one date, and one Coding Plan route.",
    "The samples do not establish an SLA, universal model latency, or a provider ranking.",
    "The client cannot separate DNS, TCP/TLS, gateway, queue, model, and network flush time.",
    "Coding Plan is a dedicated coding-tool product rather than a general application backend.",
    "Concurrent requests are a bounded smoke load, not a capacity or stress test.",
    "Buffered delivery can make client-side tokens-per-second calculations misleading."
  ]
}