{
  "schema_version": 1,
  "created_at": "2026-08-06T03:11:47.640376+00:00",
  "definition": "First place means rank 1 by both fertility and effective bits per byte among the successfully loaded, revision-pinned public comparison set.",
  "datasets": {
    "public_korean": {
      "id": "eduagarcia/multilingual_tokenizer_benchmark",
      "revision": "74fa5907af0265f42c09e21379bf21847e066990",
      "domains": [
        "finance",
        "legal",
        "lyrics",
        "news",
        "social",
        "subtitles",
        "web",
        "wiki"
      ],
      "selection_role": "development evaluation",
      "text_examples_redistributed": false,
      "results": {
        "korbyte-128k": {
          "model_id": "dawncr0w/KorByte-128K",
          "revision": null,
          "vocabulary_size": 128256,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2318164,
            "fertility": 1.4462979761995376,
            "characters_per_token": 2.6143340160575352,
            "bytes_per_token": 5.4064143002824645,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 99566,
            "ebpb": 3.07104942931184,
            "throughput_mib_s": 17.796041101582244
          },
          "ranks": {
            "fertility": 1,
            "ebpb": 1
          }
        },
        "ax-4.0": {
          "model_id": "skt/A.X-4.0",
          "revision": "a3393fd67bf09a34a85e66dfd2c2da61c4497b67",
          "vocabulary_size": 102402,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2497463,
            "fertility": 1.5581622708890421,
            "characters_per_token": 2.4266445588983703,
            "bytes_per_token": 5.018274545008274,
            "exact_document_ratio": 0.9987365761212887,
            "byte_fidelity": 0.9999967286246539,
            "character_fidelity": 0.9999976899424218,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 71975,
            "ebpb": 3.2153162310868675,
            "throughput_mib_s": 17.108109016528726
          },
          "ranks": {
            "fertility": 5,
            "ebpb": 5
          }
        },
        "trillion-7b": {
          "model_id": "trillionlabs/Trillion-7B-preview",
          "revision": "bc1650c5a954c7de0509d28c8dbc5bea560856e1",
          "vocabulary_size": 128128,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2584960,
            "fertility": 1.612751477702508,
            "characters_per_token": 2.3445062979697946,
            "bytes_per_token": 4.848413515102748,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 52099,
            "ebpb": 3.2317722100425508,
            "throughput_mib_s": 17.197271438330166
          },
          "ranks": {
            "fertility": 7,
            "ebpb": 6
          }
        },
        "midm-2.0": {
          "model_id": "K-intelligence/Midm-2.0-Base-Instruct",
          "revision": "35479c5fc9a18a5db7cc6dbadcf1db68db7beab0",
          "vocabulary_size": 131384,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2570110,
            "fertility": 1.6034865918072205,
            "characters_per_token": 2.358052768169456,
            "bytes_per_token": 4.876427468085024,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 68991,
            "ebpb": 3.2962902979186577,
            "throughput_mib_s": 17.83905852965348
          },
          "ranks": {
            "fertility": 6,
            "ebpb": 7
          }
        },
        "k-exaone": {
          "model_id": "LGAI-EXAONE/K-EXAONE-236B-A23B",
          "revision": "61e6d578eb102b578e5704e2916ac841df9eca0a",
          "vocabulary_size": 153600,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2449553,
            "fertility": 1.5282713157884886,
            "characters_per_token": 2.474106500247188,
            "bytes_per_token": 5.116425323314091,
            "exact_document_ratio": 0.9987365761212887,
            "byte_fidelity": 0.9999967286246539,
            "character_fidelity": 0.9999976899424218,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 65457,
            "ebpb": 3.1268693967808563,
            "throughput_mib_s": 28.686802012815132
          },
          "ranks": {
            "fertility": 2,
            "ebpb": 3
          }
        },
        "solar-pro3": {
          "model_id": "upstage/solar-pro3-tokenizer",
          "revision": "c41b71f0519c580610cb0fd2af4ed2b23cad544f",
          "vocabulary_size": 196608,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2453228,
            "fertility": 1.5305641410858073,
            "characters_per_token": 2.4704002237052567,
            "bytes_per_token": 5.108760783751041,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 70822,
            "ebpb": 3.1537804660058923,
            "throughput_mib_s": 18.1775053579418
          },
          "ranks": {
            "fertility": 3,
            "ebpb": 4
          }
        },
        "thunder": {
          "model_id": "thunder-research-group/Llama-Thunder-LLM-8B",
          "revision": "ecfd6c909944939ae42982f3d87cf36898da308b",
          "error": "Exception: data did not match any variant of untagged enum ModelUntagged at line 770599 column 1"
        },
        "kanana-2": {
          "model_id": "kakaocorp/kanana-2-3b-base",
          "revision": "9fa2a678c379700868d72685a22ea8ed0263ef33",
          "vocabulary_size": 128256,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2482924,
            "fertility": 1.5490914172842218,
            "characters_per_token": 2.4408540092246076,
            "bytes_per_token": 5.04765953367884,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 55554,
            "ebpb": 3.122556714986635,
            "throughput_mib_s": 24.74148599340686
          },
          "ranks": {
            "fertility": 4,
            "ebpb": 2
          }
        },
        "hyperclovax": {
          "model_id": "naver-hyperclovax/HyperCLOVAX-SEED-Think-32B",
          "revision": "a6cdfd3464d1b767259cad23e164eaf39d3e3960",
          "vocabulary_size": 128256,
          "metrics": {
            "documents": 9498,
            "words": 1602826,
            "characters": 6060455,
            "utf8_bytes": 12532955,
            "tokens": 2908889,
            "fertility": 1.8148501459297515,
            "characters_per_token": 2.08342600903644,
            "bytes_per_token": 4.308502318239025,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 53073,
            "ebpb": 3.642957417899077,
            "throughput_mib_s": 17.31447378801254
          },
          "ranks": {
            "fertility": 8,
            "ebpb": 8
          }
        }
      }
    },
    "kmmlu_test": {
      "id": "HAERAE-HUB/KMMLU",
      "revision": "d61b3f19e552c576bf5960dd24289763edc36a88",
      "subjects": 45,
      "selection_role": "frozen post-selection audit; no algorithm changes followed",
      "text_examples_redistributed": false,
      "results": {
        "korbyte-128k": {
          "model_id": "dawncr0w/KorByte-128K",
          "revision": null,
          "vocabulary_size": 128256,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2136460,
            "fertility": 1.9363674100499033,
            "characters_per_token": 2.24028345955459,
            "bytes_per_token": 4.803871825355962,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 58038,
            "ebpb": 3.2941574545664487,
            "throughput_mib_s": 15.39604399507986
          },
          "ranks": {
            "fertility": 1,
            "ebpb": 1
          }
        },
        "ax-4.0": {
          "model_id": "skt/A.X-4.0",
          "revision": "a3393fd67bf09a34a85e66dfd2c2da61c4497b67",
          "vocabulary_size": 102402,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2450486,
            "fertility": 2.220982948046557,
            "characters_per_token": 1.953194590787297,
            "bytes_per_token": 4.1882630629189475,
            "exact_document_ratio": 0.995575221238938,
            "byte_fidelity": 0.9999150369082789,
            "character_fidelity": 0.9999389922352995,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 43443,
            "ebpb": 3.6792538251829634,
            "throughput_mib_s": 14.38818129811692
          },
          "ranks": {
            "fertility": 6,
            "ebpb": 7
          }
        },
        "trillion-7b": {
          "model_id": "trillionlabs/Trillion-7B-preview",
          "revision": "bc1650c5a954c7de0509d28c8dbc5bea560856e1",
          "vocabulary_size": 128128,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2456313,
            "fertility": 2.226264213737635,
            "characters_per_token": 1.9485611157861396,
            "bytes_per_token": 4.178327436283568,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 34265,
            "ebpb": 3.6053775814055884,
            "throughput_mib_s": 14.986220352651252
          },
          "ranks": {
            "fertility": 7,
            "ebpb": 5
          }
        },
        "midm-2.0": {
          "model_id": "K-intelligence/Midm-2.0-Base-Instruct",
          "revision": "35479c5fc9a18a5db7cc6dbadcf1db68db7beab0",
          "vocabulary_size": 131384,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2450350,
            "fertility": 2.220859685281157,
            "characters_per_token": 1.953302997530965,
            "bytes_per_token": 4.188495521048013,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 42024,
            "ebpb": 3.666931429668025,
            "throughput_mib_s": 15.025529566263533
          },
          "ranks": {
            "fertility": 5,
            "ebpb": 6
          }
        },
        "k-exaone": {
          "model_id": "LGAI-EXAONE/K-EXAONE-236B-A23B",
          "revision": "61e6d578eb102b578e5704e2916ac841df9eca0a",
          "vocabulary_size": 153600,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2367509,
            "fertility": 2.145777253306796,
            "characters_per_token": 2.021650604073733,
            "bytes_per_token": 4.335054270121043,
            "exact_document_ratio": 0.995575221238938,
            "byte_fidelity": 0.9999150369082789,
            "character_fidelity": 0.9999389922352995,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 38758,
            "ebpb": 3.516715619432391,
            "throughput_mib_s": 19.086997350623356
          },
          "ranks": {
            "fertility": 3,
            "ebpb": 4
          }
        },
        "solar-pro3": {
          "model_id": "upstage/solar-pro3-tokenizer",
          "revision": "c41b71f0519c580610cb0fd2af4ed2b23cad544f",
          "vocabulary_size": 196608,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2315383,
            "fertility": 2.0985331730917385,
            "characters_per_token": 2.0671638342339045,
            "bytes_per_token": 4.432648939721852,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 48897,
            "ebpb": 3.5142549173071034,
            "throughput_mib_s": 15.146371887703577
          },
          "ranks": {
            "fertility": 2,
            "ebpb": 2
          }
        },
        "thunder": {
          "model_id": "thunder-research-group/Llama-Thunder-LLM-8B",
          "revision": "ecfd6c909944939ae42982f3d87cf36898da308b",
          "error": "Exception: data did not match any variant of untagged enum ModelUntagged at line 770599 column 1"
        },
        "kanana-2": {
          "model_id": "kakaocorp/kanana-2-3b-base",
          "revision": "9fa2a678c379700868d72685a22ea8ed0263ef33",
          "vocabulary_size": 128256,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2385998,
            "fertility": 2.1625346449941723,
            "characters_per_token": 2.0059849170032833,
            "bytes_per_token": 4.301462113547455,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 35675,
            "ebpb": 3.51569429418471,
            "throughput_mib_s": 17.819925688602993
          },
          "ranks": {
            "fertility": 4,
            "ebpb": 3
          }
        },
        "hyperclovax": {
          "model_id": "naver-hyperclovax/HyperCLOVAX-SEED-Think-32B",
          "revision": "a6cdfd3464d1b767259cad23e164eaf39d3e3960",
          "vocabulary_size": 128256,
          "metrics": {
            "documents": 35030,
            "words": 1103334,
            "characters": 4786276,
            "utf8_bytes": 10263280,
            "tokens": 2775635,
            "fertility": 2.5156797488339886,
            "characters_per_token": 1.7243895541020342,
            "bytes_per_token": 3.6976331542151617,
            "exact_document_ratio": 1.0,
            "byte_fidelity": 1.0,
            "character_fidelity": 1.0,
            "unknown_token_ratio": 0.0,
            "observed_vocabulary": 31612,
            "ebpb": 4.042635942884232,
            "throughput_mib_s": 14.72578371433208
          },
          "ranks": {
            "fertility": 8,
            "ebpb": 8
          }
        }
      }
    }
  },
  "first_place_gate_passed": true,
  "comparison_coverage_gate_passed": true,
  "required_successful_systems": [
    "ax-4.0",
    "hyperclovax",
    "k-exaone",
    "kanana-2",
    "korbyte-128k",
    "midm-2.0",
    "solar-pro3",
    "trillion-7b"
  ],
  "limitations": [
    "Intrinsic tokenization metrics do not establish downstream model quality.",
    "The comparison set is broad and current but cannot prove a universal ranking.",
    "Thunder's public tokenizer uses an unsupported custom `Beta` model and is recorded as unavailable."
  ],
  "environment": {
    "platform": "macOS-26.5-arm64-arm-64bit",
    "processor": "arm",
    "python": "3.11.15",
    "logical_cpu_count": 15,
    "physical_memory_bytes": 51539607552,
    "packages": {
      "huggingface-hub": "1.26.0",
      "rapidfuzz": "3.14.5",
      "tokenizers": "0.22.2",
      "transformers": "5.14.1"
    }
  }
}