{
  "schema": "s2s-bench-v1",
  "kind": "bill",
  "round": "opencall-r2",
  "currency": "USD",
  "figures_are": "upper bounds by construction: actual token counts from each endpoint's own counters, priced at the UNCACHED input rate",
  "how_every_line_re_derives": "each receipt below carries the token counts it was billed on and the rate they were priced at, and `usd_re_derived_from_these_tokens` is that multiplication performed: prompt × input/1e6 + completion × output/1e6. It is published per receipt and per line so the page's claim that every line re-derives is a check rather than an assurance — the judge-leg counts live in each run file's `spend` block and the arm-leg counts in each ledger row's summary, and both are joined here. A line's published total sums the round's own per-leg figures, each already rounded to four decimal places, while the re-derivation sums the unrounded products — so the two can differ in the last place, and both print rather than one being quietly conformed to the other.",
  "lines": [
    {
      "line": "kimi-k3",
      "billed_by": "carriage artifact (round-wide, every leg on this key)",
      "calls": 31,
      "usd_upper_bound": 0.989,
      "cap_usd": 18.49,
      "cap_basis": "prereg §3.4: kimi-k3 <= $18.49, stated as DERIVED headroom, with the balance probe's $0.000363 already on the bill.",
      "receipts": [
        {
          "file": "results/s1/LEDGER-S1.json",
          "usd": 0.0406,
          "calls": 5,
          "prompt_tokens": 8817,
          "completion_tokens": 941,
          "pricing_usd_per_mtok": {
            "input": 3.0,
            "cached_input": 0.3,
            "output": 15.0
          },
          "usd_re_derived_from_these_tokens": 0.040566
        },
        {
          "file": "results/s2s3/LEDGER-S2.json",
          "usd": 0.0184,
          "calls": 2,
          "prompt_tokens": 3582,
          "completion_tokens": 507,
          "pricing_usd_per_mtok": {
            "input": 3.0,
            "cached_input": 0.3,
            "output": 15.0
          },
          "usd_re_derived_from_these_tokens": 0.018351
        },
        {
          "file": "results/s2s3/LEDGER-S3.json",
          "usd": 0.0166,
          "calls": 2,
          "prompt_tokens": 3944,
          "completion_tokens": 317,
          "pricing_usd_per_mtok": {
            "input": 3.0,
            "cached_input": 0.3,
            "output": 15.0
          },
          "usd_re_derived_from_these_tokens": 0.016587
        },
        {
          "file": "results/judge/opencall-r1/runs/RUN-kimi-k3.json",
          "usd": 0.4167,
          "calls": 11,
          "prompt_tokens": 50514,
          "completion_tokens": 17676,
          "pricing_usd_per_mtok": {
            "input": 3.0,
            "cached_input": 0.3,
            "output": 15.0
          },
          "usd_re_derived_from_these_tokens": 0.416682
        },
        {
          "file": "results/judge/opencall-r2/runs/RUN-kimi-k3.json",
          "usd": 0.4667,
          "calls": 10,
          "prompt_tokens": 77285,
          "completion_tokens": 15659,
          "pricing_usd_per_mtok": {
            "input": 3.0,
            "cached_input": 0.3,
            "output": 15.0
          },
          "usd_re_derived_from_these_tokens": 0.46674
        },
        {
          "file": "results/judge/opencall-audition/AUDITION-kimi-k3.json",
          "usd": 0.03,
          "calls": 1,
          "prompt_tokens": 2716,
          "completion_tokens": 1456,
          "pricing_usd_per_mtok": {
            "input": 3.0,
            "cached_input": 0.3,
            "output": 15.0
          },
          "usd_re_derived_from_these_tokens": 0.029988
        }
      ],
      "prompt_tokens": 146858,
      "completion_tokens": 36556,
      "pricing_usd_per_mtok": {
        "cached_input": 0.3,
        "input": 3.0,
        "output": 15.0
      },
      "usd_re_derived_from_these_tokens": 0.9889
    },
    {
      "line": "gpt-5.5-2026-04-23",
      "billed_by": "carriage artifact (round-wide, every leg on this key)",
      "calls": 16,
      "usd_upper_bound": 1.2757,
      "cap_usd": 20.0,
      "cap_basis": "prereg §3.4: OpenAI <= $20.00 across the key, from the cited rates against measured tokens.",
      "receipts": [
        {
          "file": "results/s1/LEDGER-S1.json",
          "usd": 0.132,
          "calls": 5,
          "prompt_tokens": 8388,
          "completion_tokens": 3001,
          "pricing_usd_per_mtok": {
            "input": 5.0,
            "cached_input": 0.5,
            "output": 30.0
          },
          "usd_re_derived_from_these_tokens": 0.13197
        },
        {
          "file": "results/s2s3/LEDGER-S2.json",
          "usd": 0.0606,
          "calls": 2,
          "prompt_tokens": 3406,
          "completion_tokens": 1453,
          "pricing_usd_per_mtok": {
            "input": 5.0,
            "cached_input": 0.5,
            "output": 30.0
          },
          "usd_re_derived_from_these_tokens": 0.06062
        },
        {
          "file": "results/s2s3/LEDGER-S3.json",
          "usd": 0.055,
          "calls": 2,
          "prompt_tokens": 3772,
          "completion_tokens": 1205,
          "pricing_usd_per_mtok": {
            "input": 5.0,
            "cached_input": 0.5,
            "output": 30.0
          },
          "usd_re_derived_from_these_tokens": 0.05501
        },
        {
          "file": "results/judge/opencall-r2/runs/RUN-openai-flagship.json",
          "usd": 0.9694,
          "calls": 6,
          "prompt_tokens": 41784,
          "completion_tokens": 25350,
          "pricing_usd_per_mtok": {
            "input": 5.0,
            "cached_input": 0.5,
            "output": 30.0
          },
          "usd_re_derived_from_these_tokens": 0.96942
        },
        {
          "file": "results/judge/opencall-audition/AUDITION-openai-flagship.json",
          "usd": 0.0587,
          "calls": 1,
          "prompt_tokens": 2645,
          "completion_tokens": 1516,
          "pricing_usd_per_mtok": {
            "input": 5.0,
            "cached_input": 0.5,
            "output": 30.0
          },
          "usd_re_derived_from_these_tokens": 0.058705
        }
      ],
      "prompt_tokens": 59995,
      "completion_tokens": 32525,
      "pricing_usd_per_mtok": {
        "cached_input": 0.5,
        "input": 5.0,
        "output": 30.0
      },
      "usd_re_derived_from_these_tokens": 1.2757
    },
    {
      "line": "openai-gpt-5.4-mini-2026-03-17",
      "billed_by": "arm-leg ledgers (this model seats no judge chair, so the carriage table — which bills by judge-seat key — carries no line for it; these are the whole of its metered spend)",
      "calls": 9,
      "usd_upper_bound": 0.0179,
      "cap_usd": null,
      "cap_basis": "shares the OpenAI key's $20.00 registered cap with the flagship line",
      "receipts": [
        {
          "leg": "S1",
          "calls": 5,
          "usd_upper_bound": 0.0099,
          "prompt_tokens": 8388,
          "completion_tokens": 810,
          "pricing_usd_per_mtok": {
            "input": 0.75,
            "cached_input": 0.075,
            "output": 4.5
          },
          "usd_re_derived_from_these_tokens": 0.009936
        },
        {
          "leg": "S2",
          "calls": 2,
          "usd_upper_bound": 0.004,
          "prompt_tokens": 3406,
          "completion_tokens": 312,
          "pricing_usd_per_mtok": {
            "input": 0.75,
            "cached_input": 0.075,
            "output": 4.5
          },
          "usd_re_derived_from_these_tokens": 0.003959
        },
        {
          "leg": "S3",
          "calls": 2,
          "usd_upper_bound": 0.004,
          "prompt_tokens": 3772,
          "completion_tokens": 268,
          "pricing_usd_per_mtok": {
            "input": 0.75,
            "cached_input": 0.075,
            "output": 4.5
          },
          "usd_re_derived_from_these_tokens": 0.004035
        }
      ],
      "prompt_tokens": 15566,
      "completion_tokens": 1390,
      "pricing_usd_per_mtok": {
        "cached_input": 0.075,
        "input": 0.75,
        "output": 4.5
      },
      "usd_re_derived_from_these_tokens": 0.0179
    }
  ],
  "discarded_warmups": {
    "registered_rule": "- **One warmup call per arm, run and DISCARDED, kept as the cold-start receipt** (`tag: \"warmup-DISCARDED\"`). The first call to a cold model is an outlier, and at n=1 that outlier IS the published latency.",
    "billing_rule": "prereg §3.4: warmups, probes, auditions and judging all appear on the bill as their own legs",
    "per_metered_arm_per_leg": 1,
    "calls": 9,
    "usd_upper_bound": 0.1162,
    "note": "these calls are BILLED and RECEIPTED and were never scored, never published as a reply, and never entered a latency figure. They are the difference between a metered arm's call count on this bill (5 in the kid's act, 2 in each of the others) and its published replies (4, 1, 1).",
    "rows": [
      {
        "arm": "cloud-kimi-k3",
        "leg": "S1",
        "label": "warmup-DISCARDED | S1-S1-ask-A",
        "usd": 0.008337,
        "prompt_tokens": 1649,
        "completion_tokens": 226
      },
      {
        "arm": "openai-gpt-5.5-2026-04-23",
        "leg": "S1",
        "label": "warmup-DISCARDED | S1-S1-ask-A",
        "usd": 0.02497,
        "prompt_tokens": 1562,
        "completion_tokens": 572
      },
      {
        "arm": "openai-gpt-5.4-mini-2026-03-17",
        "leg": "S1",
        "label": "warmup-DISCARDED | S1-S1-ask-A",
        "usd": 0.002166,
        "prompt_tokens": 1562,
        "completion_tokens": 221
      },
      {
        "arm": "cloud-kimi-k3",
        "leg": "S2",
        "label": "warmup-DISCARDED | S2-S2-ask-A",
        "usd": 0.009528,
        "prompt_tokens": 1791,
        "completion_tokens": 277
      },
      {
        "arm": "openai-gpt-5.5-2026-04-23",
        "leg": "S2",
        "label": "warmup-DISCARDED | S2-S2-ask-A",
        "usd": 0.030205,
        "prompt_tokens": 1703,
        "completion_tokens": 723
      },
      {
        "arm": "openai-gpt-5.4-mini-2026-03-17",
        "leg": "S2",
        "label": "warmup-DISCARDED | S2-S2-ask-A",
        "usd": 0.00193,
        "prompt_tokens": 1703,
        "completion_tokens": 145
      },
      {
        "arm": "cloud-kimi-k3",
        "leg": "S3",
        "label": "warmup-DISCARDED | S3-S3-ask-A",
        "usd": 0.008196,
        "prompt_tokens": 1972,
        "completion_tokens": 152
      },
      {
        "arm": "openai-gpt-5.5-2026-04-23",
        "leg": "S3",
        "label": "warmup-DISCARDED | S3-S3-ask-A",
        "usd": 0.02884,
        "prompt_tokens": 1886,
        "completion_tokens": 647
      },
      {
        "arm": "openai-gpt-5.4-mini-2026-03-17",
        "leg": "S3",
        "label": "warmup-DISCARDED | S3-S3-ask-A",
        "usd": 0.001986,
        "prompt_tokens": 1886,
        "completion_tokens": 127
      }
    ]
  },
  "sealed_captures": {
    "by_role": {
      "judged": [
        {
          "capture": "S1-S1-ask-A",
          "replies_recorded": 40
        },
        {
          "capture": "S1-S1-ask-B",
          "replies_recorded": 40
        },
        {
          "capture": "S2-S2-ask-A",
          "replies_recorded": 20
        },
        {
          "capture": "S3-S3-ask-A",
          "replies_recorded": 20
        }
      ],
      "reshoot-spare": [
        {
          "capture": "S2-S2-ask-B",
          "replies_recorded": 0
        },
        {
          "capture": "S3-S3-ask-B",
          "replies_recorded": 0
        }
      ]
    },
    "note": "the judged captures are the four asks every arm answered — 120 replies between them. The reshoot-spare captures were sealed in the same pass as registered headroom, in case a judged ask had to be re-shot; the reply counts above are read from the arms' own record files, and both spares read zero: no arm was ever run against them and they carry no cell in the scores. They are hashed in seal-manifest.json like every other capture, which is why a reader can see they exist at all."
  },
  "metered_total_usd_upper_bound": 2.2826,
  "metered_calls": 56,
  "caps_usd": 38.49,
  "reconciliation": {
    "carriage_artifact_total_usd": 2.2647,
    "carriage_artifact_calls": 47,
    "difference_usd": 0.0179,
    "difference_calls": 9,
    "why": "the carriage artifact bills by judge-seat key. `openai-gpt-5.4-mini-2026-03-17` is a metered ARM that seats no judge chair, so its receipted arm-leg dollars were never joined to a carriage line. The page publishes the joined total and names the gap rather than quoting the smaller figure."
  },
  "cost_states": {
    "figure": "metered, priced at a cited rate against measured tokens",
    "$0.00*": "plan-included: no per-token charge on this key, but a real, unitemisable subscription sits behind the zero",
    "$0.00 local": "answered from a local daemon, where no bill exists to read",
    "em dash": "no cost figure held AND no token counters either — the three agent-transport arms. Never a zero, because a zero would read as free."
  },
  "ledger_lines_by_arm": {
    "cloud-kimi-k3": [
      {
        "leg": "S1",
        "calls": 5,
        "usd_upper_bound": 0.0406,
        "prompt_tokens": 8817,
        "completion_tokens": 941,
        "pricing_usd_per_mtok": {
          "input": 3.0,
          "cached_input": 0.3,
          "output": 15.0
        },
        "usd_re_derived_from_these_tokens": 0.040566
      },
      {
        "leg": "S2",
        "calls": 2,
        "usd_upper_bound": 0.0184,
        "prompt_tokens": 3582,
        "completion_tokens": 507,
        "pricing_usd_per_mtok": {
          "input": 3.0,
          "cached_input": 0.3,
          "output": 15.0
        },
        "usd_re_derived_from_these_tokens": 0.018351
      },
      {
        "leg": "S3",
        "calls": 2,
        "usd_upper_bound": 0.0166,
        "prompt_tokens": 3944,
        "completion_tokens": 317,
        "pricing_usd_per_mtok": {
          "input": 3.0,
          "cached_input": 0.3,
          "output": 15.0
        },
        "usd_re_derived_from_these_tokens": 0.016587
      }
    ],
    "openai-gpt-5.5-2026-04-23": [
      {
        "leg": "S1",
        "calls": 5,
        "usd_upper_bound": 0.132,
        "prompt_tokens": 8388,
        "completion_tokens": 3001,
        "pricing_usd_per_mtok": {
          "input": 5.0,
          "cached_input": 0.5,
          "output": 30.0
        },
        "usd_re_derived_from_these_tokens": 0.13197
      },
      {
        "leg": "S2",
        "calls": 2,
        "usd_upper_bound": 0.0606,
        "prompt_tokens": 3406,
        "completion_tokens": 1453,
        "pricing_usd_per_mtok": {
          "input": 5.0,
          "cached_input": 0.5,
          "output": 30.0
        },
        "usd_re_derived_from_these_tokens": 0.06062
      },
      {
        "leg": "S3",
        "calls": 2,
        "usd_upper_bound": 0.055,
        "prompt_tokens": 3772,
        "completion_tokens": 1205,
        "pricing_usd_per_mtok": {
          "input": 5.0,
          "cached_input": 0.5,
          "output": 30.0
        },
        "usd_re_derived_from_these_tokens": 0.05501
      }
    ],
    "openai-gpt-5.4-mini-2026-03-17": [
      {
        "leg": "S1",
        "calls": 5,
        "usd_upper_bound": 0.0099,
        "prompt_tokens": 8388,
        "completion_tokens": 810,
        "pricing_usd_per_mtok": {
          "input": 0.75,
          "cached_input": 0.075,
          "output": 4.5
        },
        "usd_re_derived_from_these_tokens": 0.009936
      },
      {
        "leg": "S2",
        "calls": 2,
        "usd_upper_bound": 0.004,
        "prompt_tokens": 3406,
        "completion_tokens": 312,
        "pricing_usd_per_mtok": {
          "input": 0.75,
          "cached_input": 0.075,
          "output": 4.5
        },
        "usd_re_derived_from_these_tokens": 0.003959
      },
      {
        "leg": "S3",
        "calls": 2,
        "usd_upper_bound": 0.004,
        "prompt_tokens": 3772,
        "completion_tokens": 268,
        "pricing_usd_per_mtok": {
          "input": 0.75,
          "cached_input": 0.075,
          "output": 4.5
        },
        "usd_re_derived_from_these_tokens": 0.004035
      }
    ]
  }
}