{ "measured_at": "2026-10-18T16:54:48+09:00", "measured_date": "2026-10-18", "script": "scripts/bench_keyless_topics.py", "script_sha256": "ef85cab1f4cae6672296fb6766662b4b5182997d6bf5c86264ec1becd088d25e", "environment": { "python": "3.14.8", "platform": "macOS-27.0.1-arm64-arm-64bit-Mach-O", "packages": { "tiktoken": "0.14.0", "google-genai": "2.31.0", "sentencepiece": "0.2.2", "protobuf": "7.36.2" } }, "tokenizers": [ { "model": "gpt-4o-mini", "encoding": "o200k_base", "vocabulary_sha256": "6b1b104fbdf5ea90370ff838e90173c0c01774c47a5ef98c470fb7375d65f386", "mapping_sha256": "600f26902d1cf6a1a5f54e37be988b3e0d911f1ff17ba7060bb361f9b5295521", "source_url": "https://github.com/openai/tiktoken/blob/main/tiktoken/model.py" }, { "model": "gemini-2.5-pro", "encoding": "gemma3", "experimental": true, "vocabulary_sha256": "1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c", "mapping_sha256": "eff7d8520c4656655cd42d3a7e981093fbd9cee8017db255c58df5112174df3c", "source_url": "https://github.com/googleapis/python-genai/blob/main/google/genai/_local_tokenizer_loader.py" } ], "api_requests": 0, "inference_requests": 0, "experiment": "same-japanese-input-two-public-tokenizers", "samples": [ { "id": "meeting", "input": { "text": "{\"task\":\"会議メモの整理\",\"audience\":\"欠席した同僚\",\"format\":\"決定事項・担当者・期限の表\",\"source\":\"新商品ページは金曜までに公開する。原稿は佐藤、確認は田中が担当。次回会議は来週月曜。\"}", "text_sha256": "b2944448632cd6f78b222e0669725fbe565cde12578df537fbc7011c7002341b", "characters": 117 }, "openai": { "text": "{\"task\":\"会議メモの整理\",\"audience\":\"欠席した同僚\",\"format\":\"決定事項・担当者・期限の表\",\"source\":\"新商品ページは金曜までに公開する。原稿は佐藤、確認は田中が担当。次回会議は来週月曜。\"}", "text_sha256": "b2944448632cd6f78b222e0669725fbe565cde12578df537fbc7011c7002341b", "characters": 117, "token_ids": [ 10848, 15921, 7534, 6016, 100258, 14141, 37938, 3385, 182917, 4294, 47070, 1963, 7534, 175993, 51501, 23085, 9553, 19169, 248, 4294, 4078, 7534, 78755, 8745, 191792, 4156, 151234, 8137, 4156, 152341, 3385, 9314, 4294, 4935, 7534, 3711, 26198, 80665, 5205, 5314, 63754, 54342, 5280, 161600, 39065, 788, 14085, 29480, 5205, 127139, 109496, 1395, 114862, 5205, 30814, 1404, 6632, 151234, 788, 11445, 10154, 6016, 100258, 5205, 6727, 56375, 3181, 63754, 788, 18583 ], "token_count": 70 }, "google": { "token_ids": [ 14937, 8071, 12375, 106265, 105233, 236945, 56980, 4337, 152303, 12375, 240757, 239521, 3547, 237432, 243602, 4337, 5939, 12375, 52509, 78285, 237322, 64764, 237457, 237322, 87680, 236945, 237467, 4337, 6265, 12375, 237365, 12753, 42042, 237048, 237662, 240480, 97569, 60710, 4042, 236924, 237728, 240193, 237048, 191919, 236951, 26985, 237048, 184836, 237053, 64764, 236924, 157811, 106265, 237048, 237282, 239594, 237401, 240480, 236924, 25938 ], "token_count": 60 } }, { "id": "mail", "input": { "text": "{\"task\":\"取引先への確認メールの下書き\",\"audience\":\"発注担当者\",\"format\":\"件名と本文、未確定事項は質問にする\",\"source\":\"納品予定は来月上旬。正確な日付と送り先を確認したい。数量や値段を推測して補わない。\"}", "text_sha256": "c11ce01842d83cdde03abf4003c6122c707f064918d1fefaf5f97f310cdc0290", "characters": 126 }, "openai": { "text": "{\"task\":\"取引先への確認メールの下書き\",\"audience\":\"発注担当者\",\"format\":\"件名と本文、未確定事項は質問にする\",\"source\":\"納品予定は来月上旬。正確な日付と送り先を確認したい。数量や値段を推測して補わない。\"}", "text_sha256": "c11ce01842d83cdde03abf4003c6122c707f064918d1fefaf5f97f310cdc0290", "characters": 126, "token_ids": [ 10848, 15921, 7534, 11001, 24371, 12231, 107871, 114862, 134549, 3385, 3771, 33528, 11852, 4294, 47070, 1963, 7534, 33962, 4855, 151234, 8137, 4294, 4078, 7534, 6095, 6224, 5330, 66491, 1395, 16392, 60097, 8745, 191792, 5205, 193023, 5280, 39065, 4294, 4935, 7534, 90840, 4328, 175525, 5205, 6727, 3181, 4286, 115023, 788, 10170, 60097, 5784, 2292, 20166, 5330, 116995, 12231, 7277, 114862, 23085, 3826, 788, 80012, 19160, 69555, 27561, 7277, 11800, 134925, 22440, 157713, 20473, 47592, 788, 18583 ], "token_count": 75 }, "google": { "token_ids": [ 14937, 8071, 12375, 96333, 237804, 28915, 26985, 61311, 102847, 60147, 4337, 152303, 12375, 238055, 237882, 64764, 237457, 4337, 5939, 12375, 237523, 237518, 237032, 56547, 236951, 238195, 89919, 78285, 237048, 70622, 74166, 4337, 6265, 12375, 239206, 237344, 53134, 237048, 237282, 237401, 237152, 241673, 236924, 89168, 237050, 179340, 237032, 151742, 237804, 94902, 42291, 236924, 39775, 237512, 119951, 237051, 238206, 239525, 3335, 239823, 94262, 236924, 25938 ], "token_count": 63 } }, { "id": "classify", "input": { "text": "{\"task\":\"問い合わせの分類\",\"audience\":\"窓口担当者\",\"format\":\"分類・根拠をJSONで返す\",\"source\":\"商品がまだ届きません。注文番号はABCDです。配送に関する問い合わせとして整理したい。\"}", "text_sha256": "6615bdbbf12ff40158041316d70ae0db693c58a510d6bae42e9af2988c0feb7a", "characters": 118 }, "openai": { "text": "{\"task\":\"問い合わせの分類\",\"audience\":\"窓口担当者\",\"format\":\"分類・根拠をJSONで返す\",\"source\":\"商品がまだ届きません。注文番号はABCDです。配送に関する問い合わせとして整理したい。\"}", "text_sha256": "6615bdbbf12ff40158041316d70ae0db693c58a510d6bae42e9af2988c0feb7a", "characters": 118, "token_ids": [ 10848, 15921, 7534, 105097, 3385, 164637, 4294, 47070, 1963, 7534, 22868, 241, 12630, 151234, 8137, 4294, 4078, 7534, 164637, 4156, 31888, 4347, 254, 7277, 8259, 4344, 16181, 4868, 4294, 4935, 7534, 26198, 6632, 166599, 52860, 11852, 63366, 788, 150718, 86656, 5205, 2912, 8257, 15121, 788, 117797, 5280, 42366, 39065, 105097, 176315, 182917, 23085, 3826, 788, 18583 ], "token_count": 56 }, "google": { "token_ids": [ 14937, 8071, 12375, 70385, 236945, 79963, 4337, 152303, 12375, 241636, 237768, 64764, 237457, 4337, 5939, 12375, 79963, 237322, 238219, 241817, 237051, 12324, 237007, 234568, 4337, 6265, 12375, 12753, 237053, 43452, 239666, 237264, 13772, 236924, 59203, 67583, 237048, 181242, 3652, 236924, 61277, 41667, 70385, 12547, 56980, 42291, 236924, 25938 ], "token_count": 48 } } ], "limitations": [ "Only the named models/local tokenizers; Google implementation experimental.", "No Claude count, model intelligence, generation speed, quality, API billing or other models inferred." ], "unit_map": { "samples[].input.characters": "字", "samples[].openai.token_count": "トークン", "samples[].google.token_count": "トークン" } }