← Back to the living research notebook
Current Top 100 comparisons
Ongoing, non-exhaustive research; not the final target-specific allocation order. These two ranked comparisons concern potential readers under stated assumptions and potential reach per workload for one common package. They do not count people observed to have learned. Educational effect, delivery, accessibility, audience overlap and standardized physical compute remain separate parts of the allocation question.
Read the current v0.11 working paper
Read the complete HTML edition.
Read or edit the Markdown source.
Fixed-budget portfolios and package comparisons: section 7.9.
Focused learning packages: section 7.13.
Practical-business education: section 7.14.
Joint audiences and allocation under uncertainty: section 7.15.
Thai, Malay and Khmer resource findings: section 7.16.
Prerequisite-expanded package allocation: section 7.17.
Zenodo mirror for v0.11: pending. Read the current v0.11 edition on GitHub and GitHub Pages. The preceding v0.10 edition remains on Zenodo; that DOI does not identify v0.11.
Original analysis and presentation are CC BY 4.0. Included OpenStax source material retains CC BY-NC-SA 4.0 and its original licence and credits; the analysis licence does not replace or relicense that material.
1. Worldwide potential reach
The screening calculation retains 775 non-English language-and-script edition profiles; this table displays the first 100 by potential readers under stated assumptions. Where target-comparable academic-language-access evidence is missing, the same disclosed assumption is used rather than deleting or lowering the language. Values are low–point–high scenarios. Rows overlap because many people use more than one language; do not sum them as unique readers or people educated.
Download the complete machine-readable CSV.
| Position | Edition profile | Language/script | Population basis | Potential readers under stated assumptions low–point–high | Evidence status |
|---|---|---|---|---|---|
| 1 | Simplified Standard Written Chinese profile | zh (Hans) | 1,286,047,333 | 900,233,133–1,151,960,048–1,273,186,860 | Common missing-cell assumption; target-comparable evidence still needed |
| 2 | Hindi, Devanagari-script profile | hi (Deva) | 580,318,483 | 406,222,938–519,812,678–574,515,298 | Common missing-cell assumption; target-comparable evidence still needed |
| 3 | Spanish, Latin-script profile | es (Latn) | 507,599,381 | 355,319,567–454,675,495–502,523,387 | Common missing-cell assumption; target-comparable evidence still needed |
| 4 | Arabic, Arabic-script profile | ar (Arab) | 378,792,526 | 265,154,768–339,298,442–375,004,601 | Common missing-cell assumption; target-comparable evidence still needed |
| 5 | French, Latin-script profile | fr (Latn) | 332,956,351 | 233,069,446–298,241,289–329,626,787 | Common missing-cell assumption; target-comparable evidence still needed |
| 6 | Urdu, Arabic-script profile | ur (Arab) | 313,093,257 | 219,165,280–280,449,183–309,962,324 | Common missing-cell assumption; target-comparable evidence still needed |
| 7 | Bangla, Bangla-script profile | bn (Beng) | 279,893,704 | 195,925,593–250,711,118–277,094,767 | Common missing-cell assumption; target-comparable evidence still needed |
| 8 | Portuguese, Latin-script profile | pt (Latn) | 249,463,919 | 174,624,743–223,454,037–246,969,280 | Common missing-cell assumption; target-comparable evidence still needed |
| 9 | Indonesian, Latin-script profile | id (Latn) | 248,501,794 | 126,363,708–222,592,226–246,016,776 | Target-comparable evidence overrides the common missing-cell assumption |
| 10 | Russian, Cyrillic-script profile | ru (Cyrl) | 201,151,469 | 140,806,028–180,178,793–199,139,954 | Common missing-cell assumption; target-comparable evidence still needed |
| 11 | Swahili, Latin-script profile | sw (Latn) | 194,102,079 | 135,871,455–173,864,394–192,161,058 | Common missing-cell assumption; target-comparable evidence still needed |
| 12 | Punjabi, Shahmukhi/Perso-Arabic-script profile | pa (Arab) | 176,654,800 | 123,658,360–158,236,223–174,888,252 | Common missing-cell assumption; target-comparable evidence still needed |
| 13 | German, Latin-script profile | de (Latn) | 141,873,197 | 99,311,238–127,081,057–140,454,465 | Common missing-cell assumption; target-comparable evidence still needed |
| 14 | Japanese, Japanese-script profile | ja (Jpan) | 117,608,755 | 82,326,128–105,346,501–116,432,667 | Common missing-cell assumption; target-comparable evidence still needed |
| 15 | Telugu, Telugu-script profile | te (Telu) | 101,457,360 | 71,020,152–90,879,101–100,442,786 | Common missing-cell assumption; target-comparable evidence still needed |
| 16 | Western Panjabi, Arabic-script profile | lah (Arab) | 100,980,828 | 70,686,580–90,452,254–99,971,020 | Common missing-cell assumption; target-comparable evidence still needed |
| 17 | Marathi, Devanagari-script profile | mr (Deva) | 98,639,100 | 69,047,370–88,354,681–97,652,709 | Common missing-cell assumption; target-comparable evidence still needed |
| 18 | Javanese, Latin-script profile | jv (Latn) | 96,145,858 | 67,302,101–86,121,393–95,184,399 | Common missing-cell assumption; target-comparable evidence still needed |
| 19 | Vietnamese, Latin-script profile | vi (Latn) | 92,370,781 | 64,659,547–82,739,917–91,447,073 | Common missing-cell assumption; target-comparable evidence still needed |
| 20 | Tamil, Tamil-script profile | ta (Taml) | 90,608,137 | 63,425,696–81,161,052–89,702,056 | Common missing-cell assumption; target-comparable evidence still needed |
| 21 | Persian, Arabic-script profile | fa (Arab) | 89,208,446 | 62,445,912–79,907,297–88,316,362 | Common missing-cell assumption; target-comparable evidence still needed |
| 22 | Wu Chinese, Simplified-script profile | wuu (Hans) | 84,962,400 | 59,473,680–76,103,957–84,112,776 | Common missing-cell assumption; target-comparable evidence still needed |
| 23 | Turkish, Latin-script profile | tr (Latn) | 82,419,542 | 57,693,679–73,826,225–81,595,347 | Common missing-cell assumption; target-comparable evidence still needed |
| 24 | Korean, Korean-script profile | ko (Kore) | 79,278,717 | 55,495,102–71,012,872–78,485,930 | Common missing-cell assumption; target-comparable evidence still needed |
| 25 | Cantonese, Simplified-script profile | yue (Hans) | 73,634,080 | 51,543,856–65,956,762–72,897,739 | Common missing-cell assumption; target-comparable evidence still needed |
| 26 | Filipino, Latin-script profile | fil (Latn) | 73,198,318 | 51,238,823–65,566,434–72,466,335 | Common missing-cell assumption; target-comparable evidence still needed |
| 27 | Egyptian Arabic, Arabic-script profile | arz (Arab) | 71,198,080 | 49,838,656–63,774,747–70,486,099 | Common missing-cell assumption; target-comparable evidence still needed |
| 28 | Italian, Latin-script profile | it (Latn) | 70,475,319 | 49,332,723–63,127,344–69,770,566 | Common missing-cell assumption; target-comparable evidence still needed |
| 29 | Gujarati, Gujarati-script profile | gu (Gujr) | 65,626,183 | 45,938,328–58,783,794–64,969,921 | Common missing-cell assumption; target-comparable evidence still needed |
| 30 | Pashto, Arabic-script profile | ps (Arab) | 58,062,881 | 40,644,017–52,009,065–57,482,252 | Common missing-cell assumption; target-comparable evidence still needed |
| 31 | Thai, Thai-script profile | th (Thai) | 55,936,800 | 39,155,760–50,104,656–55,377,432 | Common missing-cell assumption; target-comparable evidence still needed |
| 32 | Kannada, Kannada-script profile | kn (Knda) | 52,137,810 | 36,496,467–46,701,760–51,616,432 | Common missing-cell assumption; target-comparable evidence still needed |
| 33 | Nigerian Pidgin, Latin-script profile | pcm (Latn) | 49,716,870 | 34,801,809–44,533,235–49,219,701 | Common missing-cell assumption; target-comparable evidence still needed |
| 34 | Malayalam, Malayalam-script profile | ml (Mlym) | 45,922,589 | 32,145,812–41,134,557–45,463,363 | Common missing-cell assumption; target-comparable evidence still needed |
| 35 | Odia, Odia-script profile | or (Orya) | 45,092,160 | 31,564,512–40,390,712–44,641,238 | Common missing-cell assumption; target-comparable evidence still needed |
| 36 | Levantine Arabic, Arabic-script profile | apc (Arab) | 43,682,528 | 30,577,770–39,128,052–43,245,703 | Common missing-cell assumption; target-comparable evidence still needed |
| 37 | Punjabi, Gurmukhi-script profile | pa (Guru) | 42,948,380 | 30,063,866–38,470,449–42,518,896 | Common missing-cell assumption; target-comparable evidence still needed |
| 38 | Hausa, Latin-script profile | ha (Latn) | 41,875,124 | 29,312,587–37,509,094–41,456,373 | Common missing-cell assumption; target-comparable evidence still needed |
| 39 | Polish, Latin-script profile | pl (Latn) | 41,500,501 | 29,050,351–37,173,530–41,085,496 | Common missing-cell assumption; target-comparable evidence still needed |
| 40 | Xiang Chinese, Simplified-script profile | hsn (Hans) | 41,065,160 | 28,745,612–36,783,579–40,654,508 | Common missing-cell assumption; target-comparable evidence still needed |
| 41 | Amharic, Ethiopic-script profile | am (Ethi) | 39,176,975 | 27,423,882–35,092,262–38,785,205 | Common missing-cell assumption; target-comparable evidence still needed |
| 42 | Traditional Written Chinese profile | zh (Hant) | 39,078,482 | 27,354,937–35,004,038–38,687,697 | Common missing-cell assumption; target-comparable evidence still needed |
| 43 | Algerian Arabic, Arabic-script profile | arq (Arab) | 39,028,675 | 27,320,072–34,959,424–38,638,388 | Common missing-cell assumption; target-comparable evidence still needed |
| 44 | Oromo, Latin-script profile | om (Latn) | 38,264,431 | 26,785,102–34,274,863–37,881,787 | Common missing-cell assumption; target-comparable evidence still needed |
| 45 | Sindhi, Arabic-script profile | sd (Arab) | 37,854,600 | 26,498,220–33,907,762–37,476,054 | Common missing-cell assumption; target-comparable evidence still needed |
| 46 | Burmese, Myanmar-script profile | my (Mymr) | 37,171,608 | 26,020,126–33,295,981–36,799,892 | Common missing-cell assumption; target-comparable evidence still needed |
| 47 | Malay, Latin-script profile | ms (Latn) | 36,798,159 | 25,758,711–32,961,469–36,430,177 | Common missing-cell assumption; target-comparable evidence still needed |
| 48 | Bhojpuri, Devanagari-script profile | bho (Deva) | 34,639,016 | 24,247,311–31,027,445–34,292,626 | Common missing-cell assumption; target-comparable evidence still needed |
| 49 | Sundanese, Latin-script profile | su (Latn) | 33,787,440 | 23,651,208–30,264,657–33,449,566 | Common missing-cell assumption; target-comparable evidence still needed |
| 50 | Dutch, Latin-script profile | nl (Latn) | 32,854,898 | 22,998,429–29,429,344–32,526,349 | Common missing-cell assumption; target-comparable evidence still needed |
| 51 | Hakka Chinese, Simplified-script profile | hak (Hans) | 32,568,920 | 22,798,244–29,173,183–32,243,231 | Common missing-cell assumption; target-comparable evidence still needed |
| 52 | Moroccan Arabic, Arabic-script profile | ary (Arab) | 32,527,212 | 22,769,048–29,135,824–32,201,940 | Common missing-cell assumption; target-comparable evidence still needed |
| 53 | Yoruba, Latin-script profile | yo (Latn) | 31,761,816 | 22,233,271–28,450,231–31,444,198 | Common missing-cell assumption; target-comparable evidence still needed |
| 54 | Uzbek, Latin-script profile | uz (Latn) | 31,561,502 | 22,093,051–28,270,802–31,245,887 | Common missing-cell assumption; target-comparable evidence still needed |
| 55 | Sudanese Arabic, Arabic-script profile | apd (Arab) | 30,785,053 | 21,549,537–27,575,308–30,477,202 | Common missing-cell assumption; target-comparable evidence still needed |
| 56 | Igbo, Latin-script profile | ig (Latn) | 30,777,110 | 21,543,977–27,568,193–30,469,339 | Common missing-cell assumption; target-comparable evidence still needed |
| 57 | Saraiki, Arabic-script profile | skr (Arab) | 30,283,680 | 21,198,576–27,126,210–29,980,843 | Common missing-cell assumption; target-comparable evidence still needed |
| 58 | Cebuano, Latin-script profile | ceb (Latn) | 28,386,480 | 19,870,536–25,426,818–28,102,615 | Common missing-cell assumption; target-comparable evidence still needed |
| 59 | Awadhi, Devanagari-script profile | awa (Deva) | 27,458,163 | 19,220,714–24,595,290–27,183,581 | Common missing-cell assumption; target-comparable evidence still needed |
| 60 | Min Nan Chinese, Simplified-script profile | nan (Hans) | 26,904,760 | 18,833,332–24,099,586–26,635,712 | Common missing-cell assumption; target-comparable evidence still needed |
| 61 | Malagasy, Latin-script profile | mg (Latn) | 26,507,430 | 18,555,201–23,743,683–26,242,356 | Common missing-cell assumption; target-comparable evidence still needed |
| 62 | Ukrainian, Cyrillic-script profile | uk (Cyrl) | 24,080,585 | 16,856,410–21,569,869–23,839,779 | Common missing-cell assumption; target-comparable evidence still needed |
| 63 | Gan Chinese, Simplified-script profile | gan (Hans) | 24,072,680 | 16,850,876–21,562,788–23,831,953 | Common missing-cell assumption; target-comparable evidence still needed |
| 64 | Bavarian, Latin-script profile | bar (Latn) | 22,819,828 | 15,973,880–20,440,562–22,591,630 | Common missing-cell assumption; target-comparable evidence still needed |
| 65 | Azerbaijani, Arabic-script profile | az (Arab) | 22,517,134 | 15,761,994–20,169,428–22,291,963 | Common missing-cell assumption; target-comparable evidence still needed |
| 66 | Nepali, Devanagari-script profile | ne (Deva) | 21,735,357 | 15,214,750–19,469,161–21,518,003 | Common missing-cell assumption; target-comparable evidence still needed |
| 67 | Maithili, Devanagari-script profile | mai (Deva) | 20,333,024 | 14,233,117–18,213,040–20,129,694 | Common missing-cell assumption; target-comparable evidence still needed |
| 68 | Romanian, Latin-script profile | ro (Latn) | 20,043,507 | 14,030,455–17,953,709–19,843,072 | Common missing-cell assumption; target-comparable evidence still needed |
| 69 | Nyanja, Latin-script profile | ny (Latn) | 18,855,692 | 13,198,984–16,889,739–18,667,135 | Common missing-cell assumption; target-comparable evidence still needed |
| 70 | Somali, Latin-script profile | so (Latn) | 18,630,626 | 13,041,438–16,688,139–18,444,320 | Common missing-cell assumption; target-comparable evidence still needed |
| 71 | Assamese, Bangla-script profile | as (Beng) | 18,318,690 | 12,823,083–16,408,727–18,135,503 | Common missing-cell assumption; target-comparable evidence still needed |
| 72 | Madurese, Latin-script profile | mad (Latn) | 17,738,406 | 12,416,884–15,888,945–17,561,022 | Common missing-cell assumption; target-comparable evidence still needed |
| 73 | Rangpuri, Bangla-script profile | rkt (Beng) | 17,024,564 | 11,917,195–15,249,530–16,854,318 | Common missing-cell assumption; target-comparable evidence still needed |
| 74 | Haryanvi, Devanagari-script profile | bgc (Deva) | 16,909,560 | 11,836,692–15,146,517–16,740,464 | Common missing-cell assumption; target-comparable evidence still needed |
| 75 | Magahi, Devanagari-script profile | mag (Deva) | 16,909,560 | 11,836,692–15,146,517–16,740,464 | Common missing-cell assumption; target-comparable evidence still needed |
| 76 | Marwari, Devanagari-script profile | mwr (Deva) | 16,909,560 | 11,836,692–15,146,517–16,740,464 | Common missing-cell assumption; target-comparable evidence still needed |
| 77 | Northeastern Thai, Thai-script profile | tts (Thai) | 16,781,040 | 11,746,728–15,031,397–16,613,230 | Common missing-cell assumption; target-comparable evidence still needed |
| 78 | Nigerian Fulfulde, Latin-script profile | fuv (Latn) | 15,862,049 | 11,103,434–14,208,223–15,703,429 | Common missing-cell assumption; target-comparable evidence still needed |
| 79 | Chhattisgarhi, Devanagari-script profile | hne (Deva) | 15,500,430 | 10,850,301–13,884,307–15,345,426 | Common missing-cell assumption; target-comparable evidence still needed |
| 80 | Khmer, Khmer-script profile | km (Khmr) | 15,186,693 | 10,630,685–13,603,281–15,034,826 | Common missing-cell assumption; target-comparable evidence still needed |
| 81 | Zulu, Latin-script profile | zu (Latn) | 14,969,091 | 10,478,364–13,408,367–14,819,400 | Common missing-cell assumption; target-comparable evidence still needed |
| 82 | Sinhala, Sinhala-script profile | si (Sinh) | 14,948,168 | 10,463,718–13,389,626–14,798,686 | Common missing-cell assumption; target-comparable evidence still needed |
| 83 | Deccan, Arabic-script profile | dcc (Arab) | 13,950,387 | 9,765,271–12,495,876–13,810,883 | Common missing-cell assumption; target-comparable evidence still needed |
| 84 | Shona, Latin-script profile | sn (Latn) | 13,891,824 | 9,724,277–12,443,419–13,752,906 | Common missing-cell assumption; target-comparable evidence still needed |
| 85 | Akan, Latin-script profile | ak (Latn) | 13,489,749 | 9,442,824–12,083,266–13,354,852 | Common missing-cell assumption; target-comparable evidence still needed |
| 86 | Min Nan Chinese, Traditional-script profile | nan (Hant) | 13,473,165 | 9,431,216–12,068,411–13,338,433 | Common missing-cell assumption; target-comparable evidence still needed |
| 87 | Swedish, Latin-script profile | sv (Latn) | 13,338,377 | 9,336,864–11,947,676–13,204,993 | Common missing-cell assumption; target-comparable evidence still needed |
| 88 | Czech, Latin-script profile | cs (Latn) | 13,236,058 | 9,265,241–11,856,026–13,103,697 | Common missing-cell assumption; target-comparable evidence still needed |
| 89 | Wolof, Latin-script profile | wo (Latn) | 13,203,204 | 9,242,243–11,826,597–13,071,172 | Common missing-cell assumption; target-comparable evidence still needed |
| 90 | Kazakh, Cyrillic-script profile | kk (Cyrl) | 12,966,989 | 9,076,892–11,615,010–12,837,319 | Common missing-cell assumption; target-comparable evidence still needed |
| 91 | Hungarian, Latin-script profile | hu (Latn) | 12,313,192 | 8,619,234–11,029,380–12,190,060 | Common missing-cell assumption; target-comparable evidence still needed |
| 92 | Greek, Greek-script profile | el (Grek) | 12,249,402 | 8,574,581–10,972,241–12,126,908 | Common missing-cell assumption; target-comparable evidence still needed |
| 93 | Low German, Latin-script profile | nds (Latn) | 12,049,256 | 8,434,479–10,792,963–11,928,763 | Common missing-cell assumption; target-comparable evidence still needed |
| 94 | Kinyarwanda, Latin-script profile | rw (Latn) | 11,963,415 | 8,374,390–10,716,072–11,843,781 | Common missing-cell assumption; target-comparable evidence still needed |
| 95 | Quechua, Latin-script profile | qu (Latn) | 11,942,570 | 8,359,799–10,697,401–11,823,144 | Common missing-cell assumption; target-comparable evidence still needed |
| 96 | Iloko, Latin-script profile | ilo (Latn) | 11,354,592 | 7,948,214–10,170,727–11,241,046 | Common missing-cell assumption; target-comparable evidence still needed |
| 97 | Luba-Lulua, Latin-script profile | lua (Latn) | 11,078,688 | 7,755,082–9,923,590–10,967,901 | Common missing-cell assumption; target-comparable evidence still needed |
| 98 | Tigrinya, Ethiopic-script profile | ti (Ethi) | 10,929,719 | 7,650,803–9,790,153–10,820,422 | Common missing-cell assumption; target-comparable evidence still needed |
| 99 | Xhosa, Latin-script profile | xh (Latn) | 10,901,721 | 7,631,205–9,765,074–10,792,704 | Common missing-cell assumption; target-comparable evidence still needed |
| 100 | Tunisian Arabic, Arabic-script profile | aeb (Arab) | 10,843,920 | 7,590,744–9,713,299–10,735,481 | Common missing-cell assumption; target-comparable evidence still needed |
2. Potential reach per workload for one educational package
Every row is assigned the same 100,000-source-token educational package and nominal QA target. The denominator is a fresh-equivalent token-event workload scenario, not measured GPU time, energy, money, or final production cost. A directly measured language-script tokenization profile is used where available; otherwise the common language-blind envelope keeps the edition visible.
Download the complete machine-readable CSV.
| Position | Edition profile | Reach position | Potential readers under stated assumptions, point | Token-event workload low–point–high | Potential reach per million workload units, point | Workload evidence |
|---|---|---|---|---|---|---|
| 1 | Simplified Standard Written Chinese profile | 1 | 1,151,960,048 | 701,412–1,342,488–3,152,301 | 858,078,700 | Measured language-script tokenization profile |
| 2 | Hindi, Devanagari-script profile | 2 | 519,812,678 | 803,584–1,494,017–3,340,973 | 347,929,450 | Measured language-script tokenization profile |
| 3 | Spanish, Latin-script profile | 3 | 454,675,495 | 733,070–1,374,854–3,157,733 | 330,708,224 | Measured language-script tokenization profile |
| 4 | Arabic, Arabic-script profile | 4 | 339,298,442 | 770,805–1,574,846–4,044,349 | 215,448,628 | Common language-blind workload envelope |
| 5 | French, Latin-script profile | 5 | 298,241,289 | 740,118–1,399,094–3,195,035 | 213,167,471 | Measured language-script tokenization profile |
| 6 | Urdu, Arabic-script profile | 6 | 280,449,183 | 821,970–1,531,482–3,411,576 | 183,122,705 | Measured language-script tokenization profile |
| 7 | Portuguese, Latin-script profile | 8 | 223,454,037 | 701,560–1,331,473–3,075,366 | 167,824,727 | Measured language-script tokenization profile |
| 8 | Indonesian, Latin-script profile | 9 | 222,592,226 | 709,328–1,342,749–3,107,115 | 165,773,561 | Measured language-script tokenization profile |
| 9 | Bangla, Bangla-script profile | 7 | 250,711,118 | 846,506–1,559,237–3,457,903 | 160,790,872 | Measured language-script tokenization profile |
| 10 | Russian, Cyrillic-script profile | 10 | 180,178,793 | 758,212–1,423,107–3,267,809 | 126,609,405 | Measured language-script tokenization profile |
| 11 | Swahili, Latin-script profile | 11 | 173,864,394 | 770,805–1,574,846–4,044,349 | 110,400,876 | Common language-blind workload envelope |
| 12 | Punjabi, Shahmukhi/Perso-Arabic-script profile | 12 | 158,236,223 | 770,805–1,574,846–4,044,349 | 100,477,258 | Common language-blind workload envelope |
| 13 | German, Latin-script profile | 13 | 127,081,057 | 731,797–1,371,530–3,141,758 | 92,656,399 | Measured language-script tokenization profile |
| 14 | Japanese, Japanese-script profile | 14 | 105,346,501 | 827,964–1,533,971–3,430,509 | 68,675,694 | Measured language-script tokenization profile |
| 15 | Javanese, Latin-script profile | 18 | 86,121,393 | 743,194–1,399,407–3,203,651 | 61,541,346 | Measured language-script tokenization profile |
| 16 | Western Panjabi, Arabic-script profile | 16 | 90,452,254 | 770,805–1,574,846–4,044,349 | 57,435,613 | Common language-blind workload envelope |
| 17 | Vietnamese, Latin-script profile | 19 | 82,739,917 | 784,041–1,455,161–3,287,146 | 56,859,648 | Measured language-script tokenization profile |
| 18 | Marathi, Devanagari-script profile | 17 | 88,354,681 | 880,198–1,610,640–3,536,437 | 54,856,862 | Measured language-script tokenization profile |
| 19 | Telugu, Telugu-script profile | 15 | 90,879,101 | 907,083–1,660,930–3,655,838 | 54,715,793 | Measured language-script tokenization profile |
| 20 | Turkish, Latin-script profile | 23 | 73,826,225 | 757,699–1,427,806–3,257,891 | 51,706,068 | Measured language-script tokenization profile |
| 21 | Persian, Arabic-script profile | 21 | 79,907,297 | 770,805–1,574,846–4,044,349 | 50,739,748 | Common language-blind workload envelope |
| 22 | Wu Chinese, Simplified-script profile | 22 | 76,103,957 | 770,805–1,574,846–4,044,349 | 48,324,693 | Common language-blind workload envelope |
| 23 | Tamil, Tamil-script profile | 20 | 81,161,052 | 923,064–1,684,248–3,666,119 | 48,188,313 | Measured language-script tokenization profile |
| 24 | Korean, Korean-script profile | 24 | 71,012,872 | 770,805–1,574,846–4,044,349 | 45,091,943 | Common language-blind workload envelope |
| 25 | Egyptian Arabic, Arabic-script profile | 27 | 63,774,747 | 752,076–1,415,869–3,240,816 | 45,042,843 | Measured language-script tokenization profile |
| 26 | Italian, Latin-script profile | 28 | 63,127,344 | 777,696–1,445,572–3,260,690 | 43,669,442 | Measured language-script tokenization profile |
| 27 | Cantonese, Simplified-script profile | 25 | 65,956,762 | 770,805–1,574,846–4,044,349 | 41,881,400 | Common language-blind workload envelope |
| 28 | Filipino, Latin-script profile | 26 | 65,566,434 | 770,805–1,574,846–4,044,349 | 41,633,549 | Common language-blind workload envelope |
| 29 | Gujarati, Gujarati-script profile | 29 | 58,783,794 | 863,396–1,596,197–3,523,729 | 36,827,395 | Measured language-script tokenization profile |
| 30 | Pashto, Arabic-script profile | 30 | 52,009,065 | 770,805–1,574,846–4,044,349 | 33,024,855 | Common language-blind workload envelope |
| 31 | Thai, Thai-script profile | 31 | 50,104,656 | 906,106–1,677,044–3,625,169 | 29,876,777 | Measured language-script tokenization profile |
| 32 | Nigerian Pidgin, Latin-script profile | 33 | 44,533,235 | 770,805–1,574,846–4,044,349 | 28,277,832 | Common language-blind workload envelope |
| 33 | Levantine Arabic, Arabic-script profile | 36 | 39,128,052 | 733,911–1,387,835–3,176,452 | 28,193,586 | Measured language-script tokenization profile |
| 34 | Kannada, Kannada-script profile | 32 | 46,701,760 | 932,717–1,680,106–3,650,077 | 27,796,911 | Measured language-script tokenization profile |
| 35 | Odia, Odia-script profile | 35 | 40,390,712 | 770,805–1,574,846–4,044,349 | 25,647,402 | Common language-blind workload envelope |
| 36 | Hausa, Latin-script profile | 38 | 37,509,094 | 797,090–1,483,525–3,352,331 | 25,283,771 | Measured language-script tokenization profile |
| 37 | Traditional Written Chinese profile | 42 | 35,004,038 | 753,738–1,421,541–3,261,082 | 24,624,002 | Measured language-script tokenization profile |
| 38 | Malayalam, Malayalam-script profile | 34 | 41,134,557 | 909,831–1,673,546–3,659,052 | 24,579,283 | Measured language-script tokenization profile |
| 39 | Polish, Latin-script profile | 39 | 37,173,530 | 827,928–1,524,800–3,405,046 | 24,379,279 | Measured language-script tokenization profile |
| 40 | Xiang Chinese, Simplified-script profile | 40 | 36,783,579 | 770,805–1,574,846–4,044,349 | 23,356,935 | Common language-blind workload envelope |
| 41 | Algerian Arabic, Arabic-script profile | 43 | 34,959,424 | 770,805–1,574,846–4,044,349 | 22,198,628 | Common language-blind workload envelope |
| 42 | Dutch, Latin-script profile | 50 | 29,429,344 | 702,836–1,341,774–3,108,455 | 21,933,157 | Measured language-script tokenization profile |
| 43 | Oromo, Latin-script profile | 44 | 34,274,863 | 770,805–1,574,846–4,044,349 | 21,763,944 | Common language-blind workload envelope |
| 44 | Sindhi, Arabic-script profile | 45 | 33,907,762 | 847,681–1,558,298–3,445,009 | 21,759,491 | Measured language-script tokenization profile |
| 45 | Sundanese, Latin-script profile | 49 | 30,264,657 | 757,328–1,406,246–3,195,904 | 21,521,599 | Measured language-script tokenization profile |
| 46 | Malay, Latin-script profile | 47 | 32,961,469 | 770,805–1,574,846–4,044,349 | 20,929,961 | Common language-blind workload envelope |
| 47 | Moroccan Arabic, Arabic-script profile | 52 | 29,135,824 | 767,095–1,452,150–3,285,070 | 20,063,920 | Measured language-script tokenization profile |
| 48 | Bhojpuri, Devanagari-script profile | 48 | 31,027,445 | 859,007–1,586,853–3,526,418 | 19,552,816 | Measured language-script tokenization profile |
| 49 | Punjabi, Gurmukhi-script profile | 37 | 38,470,449 | 1,107,318–1,983,723–4,167,417 | 19,393,058 | Measured language-script tokenization profile |
| 50 | Hakka Chinese, Simplified-script profile | 51 | 29,173,183 | 770,805–1,574,846–4,044,349 | 18,524,465 | Common language-blind workload envelope |
| 51 | Uzbek, Latin-script profile | 54 | 28,270,802 | 770,805–1,574,846–4,044,349 | 17,951,469 | Common language-blind workload envelope |
| 52 | Igbo, Latin-script profile | 56 | 27,568,193 | 816,016–1,536,233–3,437,229 | 17,945,322 | Measured language-script tokenization profile |
| 53 | Sudanese Arabic, Arabic-script profile | 55 | 27,575,308 | 770,805–1,574,846–4,044,349 | 17,509,842 | Common language-blind workload envelope |
| 54 | Saraiki, Arabic-script profile | 57 | 27,126,210 | 770,805–1,574,846–4,044,349 | 17,224,673 | Common language-blind workload envelope |
| 55 | Cebuano, Latin-script profile | 58 | 25,426,818 | 827,833–1,542,010–3,430,135 | 16,489,398 | Measured language-script tokenization profile |
| 56 | Yoruba, Latin-script profile | 53 | 28,450,231 | 955,850–1,753,626–3,788,350 | 16,223,657 | Measured language-script tokenization profile |
| 57 | Awadhi, Devanagari-script profile | 59 | 24,595,290 | 832,378–1,556,175–3,500,876 | 15,804,968 | Measured language-script tokenization profile |
| 58 | Min Nan Chinese, Simplified-script profile | 60 | 24,099,586 | 770,805–1,574,846–4,044,349 | 15,302,819 | Common language-blind workload envelope |
| 59 | Malagasy, Latin-script profile | 61 | 23,743,683 | 770,805–1,574,846–4,044,349 | 15,076,827 | Common language-blind workload envelope |
| 60 | Burmese, Myanmar-script profile | 46 | 33,295,981 | 1,277,664–2,238,094–4,527,655 | 14,876,938 | Measured language-script tokenization profile |
| 61 | Gan Chinese, Simplified-script profile | 63 | 21,562,788 | 770,805–1,574,846–4,044,349 | 13,691,997 | Common language-blind workload envelope |
| 62 | Ukrainian, Cyrillic-script profile | 62 | 21,569,869 | 853,060–1,576,517–3,484,223 | 13,681,980 | Measured language-script tokenization profile |
| 63 | Bavarian, Latin-script profile | 64 | 20,440,562 | 770,805–1,574,846–4,044,349 | 12,979,402 | Common language-blind workload envelope |
| 64 | Azerbaijani, Arabic-script profile | 65 | 20,169,428 | 770,805–1,574,846–4,044,349 | 12,807,237 | Common language-blind workload envelope |
| 65 | Nepali, Devanagari-script profile | 66 | 19,469,161 | 770,805–1,574,846–4,044,349 | 12,362,580 | Common language-blind workload envelope |
| 66 | Romanian, Latin-script profile | 68 | 17,953,709 | 828,369–1,521,912–3,390,580 | 11,796,814 | Measured language-script tokenization profile |
| 67 | Maithili, Devanagari-script profile | 67 | 18,213,040 | 845,459–1,580,676–3,504,506 | 11,522,314 | Measured language-script tokenization profile |
| 68 | Nyanja, Latin-script profile | 69 | 16,889,739 | 805,625–1,512,480–3,387,245 | 11,166,916 | Measured language-script tokenization profile |
| 69 | Somali, Latin-script profile | 70 | 16,688,139 | 844,296–1,564,423–3,461,868 | 10,667,282 | Measured language-script tokenization profile |
| 70 | Amharic, Ethiopic-script profile | 41 | 35,092,262 | 2,052,019–3,460,825–6,541,585 | 10,139,854 | Measured language-script tokenization profile |
| 71 | Madurese, Latin-script profile | 72 | 15,888,945 | 770,805–1,574,846–4,044,349 | 10,089,205 | Common language-blind workload envelope |
| 72 | Magahi, Devanagari-script profile | 75 | 15,146,517 | 851,336–1,561,186–3,461,111 | 9,701,929 | Measured language-script tokenization profile |
| 73 | Assamese, Bangla-script profile | 71 | 16,408,727 | 939,052–1,691,869–3,665,232 | 9,698,578 | Measured language-script tokenization profile |
| 74 | Rangpuri, Bangla-script profile | 73 | 15,249,530 | 770,805–1,574,846–4,044,349 | 9,683,187 | Common language-blind workload envelope |
| 75 | Haryanvi, Devanagari-script profile | 74 | 15,146,517 | 770,805–1,574,846–4,044,349 | 9,617,776 | Common language-blind workload envelope |
| 76 | Marwari, Devanagari-script profile | 76 | 15,146,517 | 770,805–1,574,846–4,044,349 | 9,617,776 | Common language-blind workload envelope |
| 77 | Northeastern Thai, Thai-script profile | 77 | 15,031,397 | 770,805–1,574,846–4,044,349 | 9,544,676 | Common language-blind workload envelope |
| 78 | Nigerian Fulfulde, Latin-script profile | 78 | 14,208,223 | 799,057–1,507,138–3,398,398 | 9,427,287 | Measured language-script tokenization profile |
| 79 | Chhattisgarhi, Devanagari-script profile | 79 | 13,884,307 | 846,572–1,567,155–3,464,433 | 8,859,563 | Measured language-script tokenization profile |
| 80 | Swedish, Latin-script profile | 87 | 11,947,676 | 740,976–1,390,028–3,180,795 | 8,595,279 | Measured language-script tokenization profile |
| 81 | Zulu, Latin-script profile | 81 | 13,408,367 | 853,436–1,563,205–3,443,752 | 8,577,486 | Measured language-script tokenization profile |
| 82 | Shona, Latin-script profile | 84 | 12,443,419 | 842,436–1,553,878–3,464,240 | 8,007,978 | Measured language-script tokenization profile |
| 83 | Deccan, Arabic-script profile | 83 | 12,495,876 | 770,805–1,574,846–4,044,349 | 7,934,665 | Common language-blind workload envelope |
| 84 | Czech, Latin-script profile | 88 | 11,856,026 | 811,706–1,505,293–3,363,419 | 7,876,222 | Measured language-script tokenization profile |
| 85 | Kazakh, Cyrillic-script profile | 90 | 11,615,010 | 790,304–1,500,038–3,350,232 | 7,743,142 | Measured language-script tokenization profile |
| 86 | Min Nan Chinese, Traditional-script profile | 86 | 12,068,411 | 770,805–1,574,846–4,044,349 | 7,663,232 | Common language-blind workload envelope |
| 87 | Wolof, Latin-script profile | 89 | 11,826,597 | 856,392–1,568,582–3,470,412 | 7,539,675 | Measured language-script tokenization profile |
| 88 | Akan, Latin-script profile | 85 | 12,083,266 | 924,290–1,733,928–3,853,794 | 6,968,724 | Measured language-script tokenization profile |
| 89 | Hungarian, Latin-script profile | 91 | 11,029,380 | 870,706–1,597,485–3,521,117 | 6,904,214 | Measured language-script tokenization profile |
| 90 | Kinyarwanda, Latin-script profile | 94 | 10,716,072 | 844,115–1,559,568–3,456,984 | 6,871,180 | Measured language-script tokenization profile |
| 91 | Low German, Latin-script profile | 93 | 10,792,963 | 770,805–1,574,846–4,044,349 | 6,853,344 | Common language-blind workload envelope |
| 92 | Tunisian Arabic, Arabic-script profile | 100 | 9,713,299 | 750,642–1,423,090–3,286,436 | 6,825,498 | Measured language-script tokenization profile |
| 93 | Quechua, Latin-script profile | 95 | 10,697,401 | 770,805–1,574,846–4,044,349 | 6,792,664 | Common language-blind workload envelope |
| 94 | Sinhala, Sinhala-script profile | 82 | 13,389,626 | 1,149,548–2,018,995–4,182,797 | 6,631,827 | Measured language-script tokenization profile |
| 95 | Xhosa, Latin-script profile | 99 | 9,765,074 | 822,924–1,524,156–3,397,755 | 6,406,871 | Measured language-script tokenization profile |
| 96 | Luba-Lulua, Latin-script profile | 97 | 9,923,590 | 829,810–1,553,112–3,514,734 | 6,389,488 | Measured language-script tokenization profile |
| 97 | Iloko, Latin-script profile | 96 | 10,170,727 | 883,714–1,620,524–3,549,823 | 6,276,195 | Measured language-script tokenization profile |
| 98 | Afrikaans, Latin-script profile | 105 | 8,927,062 | 758,173–1,426,135–3,228,689 | 6,259,618 | Measured language-script tokenization profile |
| 99 | Greek, Greek-script profile | 92 | 10,972,241 | 982,304–1,759,491–3,779,844 | 6,236,033 | Measured language-script tokenization profile |
| 100 | Umbundu, Latin-script profile | 101 | 9,663,755 | 858,759–1,595,101–3,530,126 | 6,058,396 | Measured language-script tokenization profile |
What remains before a defensible allocation recommendation
The study still has to compare exact educational packages at the relevant stage and subject; measure independently usable existing supply and additional learning effect; account for offline, phone, signed-language and other accessibility delivery; avoid double-counting multilingual and displaced beneficiaries; count shared source preparation and interlanguage reuse once; and convert workload into standardized physical compute. Until then, these are transparent reach and common-package workload comparisons—not a final instruction to spend money in rank order.
Living checkpoint v0.11, published 9 September 2026. Missing evidence is unfinished research, not evidence of no need.