champollion 0.3.3

package/shared/catalogue/method-coverage.json

@@ -0,0 +1,1751 @@
+{
+  "version": 2,
+  "asOf": "2026-06-28",
+  "note": "Per-method MT coverage (supported languages) imported cite-only from each provider's published list; ISO 639-3. COUNT SEMANTICS (settled 2026-08-16 — read this before quoting any number): `count` is NOT one kind of quantity, so every entry now declares `countBasis`. 'publisher-list-rows' = the length of the provider's own published list exactly as they publish it (table rows / tags), cite-only — this is the figure public prose may attribute to the provider by name. 'champollion-derived-enumeration' = OUR collapse of their list to distinct ISO 639-3 base languages — never attributable to the provider (see CLAUDE.md derived-provenance rule). 'publisher-stated-headline' = a total the publisher asserts with no enumerable list behind it. `iso6393Count` always equals len(iso6393) and is always Champollion-derived; for 'publisher-list-rows' entries it is LOWER than `count` because script/region variants collapse (Google zh-CN+zh-TW→zho, NLLB's 200 FLORES variants→196 base codes, LibreTranslate zt/pb→zho/por). That gap is arithmetic, not a defect — it was misread as one in docs/PROVIDER_COUNT_VERIFICATION_2026-08-01.md §3.1. Map coverage and tier math read the iso6393 ARRAY, never `count`. tier: 'service' = deployable consumer/API service; 'open' = open research models. anyToAny: true = the method can intertranslate within its whole set; false (OPUS-MT, Tilde) = per-pair/per-engine models only — coverage of a language does NOT imply any given pair exists. NOT reproduced by Champollion. OPUS-MT iso6393 list pending (count cited). v2 (2026-07-22): releaseDate + releaseDate_source_url added where cleanly citable (Tilde/OPUS omitted — no verifiable date); omt1600 added COUNT-ONLY (cited, never joins map coverage/tier counts) with Meta’s own >400-of-1,600 quality self-report for the omnimodel-escalation timeline. INGEST NOTE 2026-08-10: entries carrying `supersededBy` are no longer ingested into the atlas — a fetched source now carries the same coverage from the owner’s own publication — but they REMAIN here because pre-cutover runtime consumers (tier logic, the queue ranker, the website) still read this file. They are deleted with the file itself at cutover, not before: retiring data ahead of its consumers is how a working CLI breaks in the name of tidiness.",
+  "methods": [
+    {
+      "key": "google-translate",
+      "label": "Google Translate",
+      "count": 194,
+      "countBasis": "publisher-list-rows",
+      "iso6393Count": 187,
+      "tier": "service",
+      "anyToAny": true,
+      "codeSystem": "BCP-47",
+      "iso6393": [
+        "abk",
+        "ace",
+        "ach",
+        "afr",
+        "aka",
+        "alz",
+        "amh",
+        "ara",
+        "asm",
+        "awa",
+        "aym",
+        "aze",
+        "bak",
+        "bam",
+        "ban",
+        "bbc",
+        "bel",
+        "bem",
+        "ben",
+        "bew",
+        "bho",
+        "bik",
+        "bos",
+        "bre",
+        "bts",
+        "btx",
+        "bua",
+        "bul",
+        "cat",
+        "ceb",
+        "ces",
+        "cgg",
+        "chm",
+        "chv",
+        "ckb",
+        "cnh",
+        "cos",
+        "crh",
+        "crs",
+        "cym",
+        "dan",
+        "deu",
+        "din",
+        "div",
+        "doi",
+        "dov",
+        "dzo",
+        "ell",
+        "eng",
+        "epo",
+        "est",
+        "eus",
+        "ewe",
+        "fas",
+        "fij",
+        "fil",
+        "fin",
+        "fra",
+        "fry",
+        "ful",
+        "gaa",
+        "gla",
+        "gle",
+        "glg",
+        "gom",
+        "grn",
+        "guj",
+        "hat",
+        "hau",
+        "haw",
+        "heb",
+        "hil",
+        "hin",
+        "hmn",
+        "hrv",
+        "hrx",
+        "hun",
+        "hye",
+        "ibo",
+        "ilo",
+        "ind",
+        "isl",
+        "ita",
+        "jav",
+        "jpn",
+        "kan",
+        "kat",
+        "kaz",
+        "khm",
+        "kin",
+        "kir",
+        "kor",
+        "kri",
+        "ktu",
+        "kur",
+        "lao",
+        "lat",
+        "lav",
+        "lij",
+        "lim",
+        "lin",
+        "lit",
+        "lmo",
+        "ltg",
+        "ltz",
+        "lug",
+        "luo",
+        "lus",
+        "mai",
+        "mak",
+        "mal",
+        "mar",
+        "min",
+        "mkd",
+        "mlg",
+        "mlt",
+        "mni",
+        "mon",
+        "mri",
+        "msa",
+        "mya",
+        "nbl",
+        "nep",
+        "new",
+        "nld",
+        "nor",
+        "nso",
+        "nus",
+        "nya",
+        "oci",
+        "ori",
+        "orm",
+        "pag",
+        "pam",
+        "pan",
+        "pap",
+        "pol",
+        "por",
+        "pus",
+        "que",
+        "rom",
+        "ron",
+        "run",
+        "rus",
+        "sag",
+        "san",
+        "scn",
+        "shn",
+        "sin",
+        "slk",
+        "slv",
+        "smo",
+        "sna",
+        "snd",
+        "som",
+        "sot",
+        "spa",
+        "sqi",
+        "srp",
+        "ssw",
+        "sun",
+        "swa",
+        "swe",
+        "szl",
+        "tam",
+        "tat",
+        "tel",
+        "tet",
+        "tgk",
+        "tha",
+        "tir",
+        "tsn",
+        "tso",
+        "tuk",
+        "tur",
+        "uig",
+        "ukr",
+        "urd",
+        "uzb",
+        "vie",
+        "xho",
+        "yid",
+        "yor",
+        "yua",
+        "yue",
+        "zho",
+        "zul"
+      ],
+      "source_url": "https://docs.cloud.google.com/translate/docs/languages",
+      "asOf": "2026-08-16",
+      "confidence": "high",
+      "verified": "confirmed",
+      "publisherTables": {
+        "nmt": 194,
+        "translationLlm": 127,
+        "translationLlmOfficial": 113,
+        "translationLlmExperimental": 14,
+        "romanization": 14,
+        "automlCustomModel": 53,
+        "statedTotal": null,
+        "note": "Counted mechanically from the fetched page HTML on 2026-08-16 (page's own stamp: 'Last updated 2026-08-11 UTC'), by counting <tr> elements carrying <td> in each table. The NMT table ('Language' / 'ISO-639 code') has 194 rows — that is what `count: 194` has always been, and it is CONFIRMED. The Translation LLM table ('Language name' / 'Language code' / 'Official Support' / 'Experimental Support') has 127 rows, of which 113 are official and 14 experimental; those 127 rows collapse to 86 distinct base language tags because the LLM table is locale-inflated (en-AU, es-419, zh-Hant, …). The page states NO combined total. SUPERSEDES the 'Translation LLM 109 / NMT 135' table in docs/PROVIDER_COUNT_VERIFICATION_2026-08-01.md §2, which was read on 2026-08-01 against the page as stamped 2026-07-31; the page has changed since. Never quote 109 or 135 for Google — 135 is Microsoft Translator's figure."
+      },
+      "pendingFounderDecision": "Product identity (defect 3 of docs/PROVIDER_COUNT_VERIFICATION_2026-08-01.md §3) is UNRESOLVED and deliberately untouched here: `label` says 'Google Translate' (the consumer product) while source_url/nature are Google Cloud Translation. The numbers are now verified; the claim we are making about WHICH product is a founder call. Do not silently rename the label — cli/website/src/data/seam-hubs.json carries the same pairing to the public homepage.",
+      "homepage": "https://translate.google.com",
+      "nature": "Commercial cloud API (Google Cloud Translation).",
+      "releaseDate": "2006-04-28",
+      "releaseDate_source_url": "https://en.wikipedia.org/wiki/Google_Translate"
+    },
+    {
+      "key": "microsoft-translator",
+      "label": "Microsoft Translator",
+      "count": 135,
+      "countBasis": "publisher-list-rows",
+      "iso6393Count": 128,
+      "tier": "service",
+      "anyToAny": true,
+      "codeSystem": "Microsoft Translator language tags (BCP-47 / mix of ISO 639-1 two-letter, ISO 639-3 three-letter, and script/region variants like zh-Hans, sr-Cyrl, fr-ca)",
+      "iso6393": [
+        "afr",
+        "amh",
+        "ara",
+        "asm",
+        "aze",
+        "bak",
+        "ben",
+        "bho",
+        "bod",
+        "bos",
+        "brx",
+        "bul",
+        "cat",
+        "ces",
+        "ckb",
+        "cym",
+        "dan",
+        "deu",
+        "div",
+        "doi",
+        "dsb",
+        "ell",
+        "eng",
+        "est",
+        "eus",
+        "fao",
+        "fas",
+        "fij",
+        "fil",
+        "fin",
+        "fra",
+        "gle",
+        "glg",
+        "gom",
+        "guj",
+        "hat",
+        "hau",
+        "heb",
+        "hin",
+        "hne",
+        "hrv",
+        "hsb",
+        "hun",
+        "hye",
+        "ibo",
+        "ikt",
+        "iku",
+        "ind",
+        "isl",
+        "ita",
+        "jpn",
+        "kan",
+        "kas",
+        "kat",
+        "kaz",
+        "khm",
+        "kin",
+        "kir",
+        "kmr",
+        "kor",
+        "lao",
+        "lav",
+        "lin",
+        "lit",
+        "lug",
+        "lzh",
+        "mai",
+        "mal",
+        "mar",
+        "mkd",
+        "mlg",
+        "mlt",
+        "mni",
+        "mon",
+        "mri",
+        "msa",
+        "mww",
+        "mya",
+        "nep",
+        "nld",
+        "nob",
+        "nso",
+        "nya",
+        "ory",
+        "otq",
+        "pan",
+        "pol",
+        "por",
+        "prs",
+        "pus",
+        "ron",
+        "run",
+        "rus",
+        "sin",
+        "slk",
+        "slv",
+        "smo",
+        "sna",
+        "snd",
+        "som",
+        "sot",
+        "spa",
+        "sqi",
+        "srp",
+        "swa",
+        "swe",
+        "tah",
+        "tam",
+        "tat",
+        "tel",
+        "tha",
+        "tir",
+        "tlh",
+        "ton",
+        "tsn",
+        "tuk",
+        "tur",
+        "uig",
+        "ukr",
+        "urd",
+        "uzb",
+        "vie",
+        "xho",
+        "yor",
+        "yua",
+        "yue",
+        "zho",
+        "zul"
+      ],
+      "source_url": "https://learn.microsoft.com/en-us/azure/ai-services/translator/language-support",
+      "asOf": "2026-08-16",
+      "confidence": "high",
+      "verified": "confirmed",
+      "publisherTables": {
+        "textTranslation": 135,
+        "documentTranslation": 113,
+        "transliteration": 37,
+        "statedTotal": null,
+        "note": "Re-verified mechanically 2026-08-16 from the fetched page HTML: the text-translation table has exactly 135 data rows, confirming `count: 135` against its own source. The 128 in iso6393Count is our collapse of Microsoft's script/region tags (zh-Hans/zh-Hant, sr-Cyrl/sr-Latn, fr-ca, pt-pt, …) to base languages."
+      },
+      "homepage": "https://azure.microsoft.com/products/ai-services/ai-translator",
+      "nature": "Commercial cloud API (Azure AI Translator).",
+      "releaseDate": "2007-09",
+      "releaseDate_source_url": "https://www.microsoft.com/en-us/translator/blog/2007/09/24/introducing-windows-live-translator-beta/",
+      "supersededBy": "vendor-languages (the vendor’s live endpoint)"
+    },
+    {
+      "key": "deepl",
+      "label": "DeepL",
+      "count": 33,
+      "countBasis": "publisher-list-rows",
+      "iso6393Count": 33,
+      "countBasisNote": "Not mechanically re-verified 2026-08-16: developers.deepl.com renders its language list client-side, so there is no table to count from fetched HTML. Still `partially-confirmed` on the 2026-06-28 transcription — the one provider in this file whose count rests on a human read.",
+      "tier": "service",
+      "anyToAny": true,
+      "codeSystem": "BCP-47 / ISO 639-1",
+      "iso6393": [
+        "ara",
+        "bul",
+        "ces",
+        "dan",
+        "deu",
+        "ell",
+        "eng",
+        "est",
+        "fin",
+        "fra",
+        "heb",
+        "hun",
+        "ind",
+        "ita",
+        "jpn",
+        "kor",
+        "lav",
+        "lit",
+        "nld",
+        "nob",
+        "pol",
+        "por",
+        "ron",
+        "rus",
+        "slk",
+        "slv",
+        "spa",
+        "swe",
+        "tha",
+        "tur",
+        "ukr",
+        "vie",
+        "zho"
+      ],
+      "source_url": "https://developers.deepl.com/docs/getting-started/supported-languages",
+      "asOf": "2026-06-28",
+      "confidence": "medium",
+      "verified": "partially-confirmed",
+      "homepage": "https://www.deepl.com",
+      "nature": "Commercial cloud API.",
+      "releaseDate": "2017-08-28",
+      "releaseDate_source_url": "https://en.wikipedia.org/wiki/DeepL_Translator",
+      "formalityIso6393": [
+        "deu",
+        "fra",
+        "ita",
+        "jpn",
+        "nld",
+        "pol",
+        "por",
+        "por-PT",
+        "rus",
+        "spa",
+        "spa-MX"
+      ],
+      "formalitySource": "DeepL API documentation (2025) — transcribed cite-only, the same lane as the language list itself; relocated 2026-08-10 from per-card flags whose formalitySource named the same document",
+      "transcriptionNote": "2026-08-11: 'arb' corrected to 'ara'. DeepL's own docs publish \"AR (Arabic)\" — the macrolanguage tag — and transcribing it as arb (Standard Arabic, an individual member) was over-resolution: an interpretation, not what the source says. The edge lands on ara exactly as published; whether members are covered stays the four-way macrolanguage question it always was, never propagated."
+    },
+    {
+      "key": "libretranslate",
+      "label": "LibreTranslate",
+      "count": 49,
+      "countBasis": "publisher-list-rows",
+      "iso6393Count": 47,
+      "countBasisNote": "49 published codes collapse to 47 base languages: LibreTranslate-internal zt (Chinese traditional) → zho and pb (Portuguese Brazil) → por, exactly as codeSystem describes.",
+      "tier": "service",
+      "anyToAny": true,
+      "codeSystem": "LibreTranslate language codes (mostly ISO 639-1, with LibreTranslate-internal codes zt=Chinese traditional and pb=Portuguese Brazil)",
+      "iso6393": [
+        "ara",
+        "aze",
+        "ben",
+        "bul",
+        "cat",
+        "ces",
+        "dan",
+        "deu",
+        "ell",
+        "eng",
+        "epo",
+        "est",
+        "eus",
+        "fas",
+        "fin",
+        "fra",
+        "gle",
+        "glg",
+        "heb",
+        "hin",
+        "hun",
+        "ind",
+        "ita",
+        "jpn",
+        "kir",
+        "kor",
+        "lav",
+        "lit",
+        "msa",
+        "nld",
+        "nob",
+        "pol",
+        "por",
+        "ron",
+        "rus",
+        "slk",
+        "slv",
+        "spa",
+        "sqi",
+        "swe",
+        "tgl",
+        "tha",
+        "tur",
+        "ukr",
+        "urd",
+        "vie",
+        "zho"
+      ],
+      "source_url": "https://docs.libretranslate.com/guides/supported_languages/",
+      "asOf": "2026-06-28",
+      "confidence": "high",
+      "verified": "confirmed",
+      "homepage": "https://libretranslate.com",
+      "nature": "Open-source, self-hostable translation API (AGPL-3.0 server, Argos Translate models).",
+      "releaseDate": "2020-12-21",
+      "releaseDate_source_url": "https://github.com/LibreTranslate/LibreTranslate/releases/tag/v1.0.0",
+      "supersededBy": "vendor-languages (the vendor’s live endpoint)"
+    },
+    {
+      "key": "nllb",
+      "label": "NLLB-200",
+      "count": 200,
+      "countBasis": "publisher-list-rows",
+      "iso6393Count": 196,
+      "countBasisNote": "200 is the FLORES-200 lang_Script variant count Meta publishes and names the model after; 196 is our collapse to distinct ISO 639-3 base codes (a language listed in two scripts counts once). A THIRD figure, 195, appears in this entry's supersededBy string from the HF model card — three numbers for one model, which is why countBasis now exists.",
+      "tier": "open",
+      "anyToAny": true,
+      "codeSystem": "FLORES-200 (ISO 639-3 + ISO 15924 script, lang_Script); iso6393 field = ISO 639-3 base codes",
+      "iso6393": [
+        "ace",
+        "acm",
+        "acq",
+        "aeb",
+        "afr",
+        "ajp",
+        "aka",
+        "als",
+        "amh",
+        "apc",
+        "arb",
+        "ars",
+        "ary",
+        "arz",
+        "asm",
+        "ast",
+        "awa",
+        "ayr",
+        "azb",
+        "azj",
+        "bak",
+        "bam",
+        "ban",
+        "bel",
+        "bem",
+        "ben",
+        "bho",
+        "bjn",
+        "bod",
+        "bos",
+        "bug",
+        "bul",
+        "cat",
+        "ceb",
+        "ces",
+        "cjk",
+        "ckb",
+        "crh",
+        "cym",
+        "dan",
+        "deu",
+        "dik",
+        "dyu",
+        "dzo",
+        "ell",
+        "eng",
+        "epo",
+        "est",
+        "eus",
+        "ewe",
+        "fao",
+        "fij",
+        "fin",
+        "fon",
+        "fra",
+        "fur",
+        "fuv",
+        "gaz",
+        "gla",
+        "gle",
+        "glg",
+        "grn",
+        "guj",
+        "hat",
+        "hau",
+        "heb",
+        "hin",
+        "hne",
+        "hrv",
+        "hun",
+        "hye",
+        "ibo",
+        "ilo",
+        "ind",
+        "isl",
+        "ita",
+        "jav",
+        "jpn",
+        "kab",
+        "kac",
+        "kam",
+        "kan",
+        "kas",
+        "kat",
+        "kaz",
+        "kbp",
+        "kea",
+        "khk",
+        "khm",
+        "kik",
+        "kin",
+        "kir",
+        "kmb",
+        "kmr",
+        "knc",
+        "kon",
+        "kor",
+        "lao",
+        "lij",
+        "lim",
+        "lin",
+        "lit",
+        "lmo",
+        "ltg",
+        "ltz",
+        "lua",
+        "lug",
+        "luo",
+        "lus",
+        "lvs",
+        "mag",
+        "mai",
+        "mal",
+        "mar",
+        "min",
+        "mkd",
+        "mlt",
+        "mni",
+        "mos",
+        "mri",
+        "mya",
+        "nld",
+        "nno",
+        "nob",
+        "npi",
+        "nso",
+        "nus",
+        "nya",
+        "oci",
+        "ory",
+        "pag",
+        "pan",
+        "pap",
+        "pbt",
+        "pes",
+        "plt",
+        "pol",
+        "por",
+        "prs",
+        "quy",
+        "ron",
+        "run",
+        "rus",
+        "sag",
+        "san",
+        "sat",
+        "scn",
+        "shn",
+        "sin",
+        "slk",
+        "slv",
+        "smo",
+        "sna",
+        "snd",
+        "som",
+        "sot",
+        "spa",
+        "srd",
+        "srp",
+        "ssw",
+        "sun",
+        "swe",
+        "swh",
+        "szl",
+        "tam",
+        "taq",
+        "tat",
+        "tel",
+        "tgk",
+        "tgl",
+        "tha",
+        "tir",
+        "tpi",
+        "tsn",
+        "tso",
+        "tuk",
+        "tum",
+        "tur",
+        "twi",
+        "tzm",
+        "uig",
+        "ukr",
+        "umb",
+        "urd",
+        "uzn",
+        "vec",
+        "vie",
+        "war",
+        "wol",
+        "xho",
+        "ydd",
+        "yor",
+        "yue",
+        "zho",
+        "zsm",
+        "zul"
+      ],
+      "source_url": "https://github.com/facebookresearch/flores/blob/main/flores200/README.md",
+      "asOf": "2026-06-28",
+      "confidence": "high",
+      "verified": "confirmed",
+      "homepage": "https://ai.meta.com/research/no-language-left-behind/",
+      "license": "CC-BY-NC-4.0 (checkpoints) — non-commercial",
+      "nature": "Open research model (Meta) — any-to-any across 200 languages; downloadable, non-commercial license.",
+      "releaseDate": "2022-07-11",
+      "releaseDate_source_url": "https://arxiv.org/abs/2207.04672",
+      "supersededBy": "hf-models → hf:facebook/nllb-200-distilled-600M (owner’s card, same 195 languages)"
+    },
+    {
+      "key": "opus",
+      "label": "OPUS-MT",
+      "count": 336,
+      "countBasis": "champollion-derived-enumeration",
+      "iso6393Count": 336,
+      "countBasisNote": "count is OUR enumeration, not Helsinki-NLP's headline — never write \"OPUS-MT supports 336 languages\" as their claim. Their published headline is 557 (publishedHeadline), at a finer granularity that is not enumerable as distinct alpha-3 codes.",
+      "tier": "open",
+      "anyToAny": false,
+      "codeSystem": "ISO 639-3",
+      "iso6393": [
+        "abk",
+        "ady",
+        "afh",
+        "afr",
+        "ain",
+        "akl",
+        "alt",
+        "amh",
+        "ang",
+        "aoz",
+        "ara",
+        "arg",
+        "asm",
+        "ast",
+        "avk",
+        "awa",
+        "aze",
+        "bak",
+        "bal",
+        "bam",
+        "bar",
+        "bel",
+        "ben",
+        "ber",
+        "bho",
+        "bod",
+        "bre",
+        "brx",
+        "bua",
+        "bul",
+        "bvy",
+        "bzt",
+        "cat",
+        "cay",
+        "cbk",
+        "ceb",
+        "ces",
+        "cha",
+        "che",
+        "chg",
+        "chm",
+        "chn",
+        "cho",
+        "chr",
+        "chv",
+        "ckt",
+        "cor",
+        "cos",
+        "cpi",
+        "cre",
+        "crh",
+        "csb",
+        "cym",
+        "dan",
+        "deu",
+        "div",
+        "dng",
+        "drt",
+        "dsb",
+        "dtp",
+        "dws",
+        "egl",
+        "ell",
+        "emx",
+        "eng",
+        "enm",
+        "epo",
+        "est",
+        "eus",
+        "evn",
+        "ewe",
+        "ext",
+        "fao",
+        "fas",
+        "fij",
+        "fin",
+        "fkv",
+        "fra",
+        "frm",
+        "fro",
+        "frr",
+        "fry",
+        "ful",
+        "fur",
+        "gag",
+        "gbm",
+        "gcf",
+        "gil",
+        "gla",
+        "gle",
+        "glg",
+        "glv",
+        "gos",
+        "got",
+        "grc",
+        "grn",
+        "gsw",
+        "guc",
+        "guj",
+        "hat",
+        "hau",
+        "haw",
+        "hbo",
+        "hbs",
+        "heb",
+        "hif",
+        "hil",
+        "hin",
+        "hmn",
+        "hoc",
+        "hrx",
+        "hsb",
+        "hun",
+        "hye",
+        "iba",
+        "ibo",
+        "ido",
+        "iii",
+        "iku",
+        "ile",
+        "ilo",
+        "ina",
+        "isl",
+        "ita",
+        "izh",
+        "jam",
+        "jav",
+        "jbo",
+        "jdt",
+        "jpa",
+        "jpn",
+        "kaa",
+        "kab",
+        "kal",
+        "kam",
+        "kan",
+        "kat",
+        "kaz",
+        "kek",
+        "kha",
+        "khm",
+        "kin",
+        "kir",
+        "kjh",
+        "klj",
+        "kok",
+        "kom",
+        "kor",
+        "krc",
+        "krl",
+        "ksh",
+        "kum",
+        "kur",
+        "kxi",
+        "kzj",
+        "laa",
+        "lad",
+        "lah",
+        "lao",
+        "lat",
+        "lav",
+        "ldn",
+        "lfn",
+        "lij",
+        "lim",
+        "lin",
+        "lit",
+        "liv",
+        "lkt",
+        "lld",
+        "lmo",
+        "ltz",
+        "lug",
+        "lut",
+        "lzz",
+        "mad",
+        "mah",
+        "mai",
+        "mal",
+        "mar",
+        "mdf",
+        "mfe",
+        "mgm",
+        "mic",
+        "mkd",
+        "mlg",
+        "mlt",
+        "mnc",
+        "mni",
+        "mnw",
+        "moh",
+        "mon",
+        "mri",
+        "msa",
+        "mus",
+        "mvv",
+        "mwl",
+        "mya",
+        "myv",
+        "nah",
+        "nau",
+        "nav",
+        "nch",
+        "nds",
+        "nep",
+        "ngt",
+        "ngu",
+        "niu",
+        "nld",
+        "nlv",
+        "nog",
+        "non",
+        "nor",
+        "nov",
+        "nst",
+        "nus",
+        "nya",
+        "nys",
+        "oar",
+        "oci",
+        "ofs",
+        "ood",
+        "ori",
+        "orv",
+        "osp",
+        "oss",
+        "ota",
+        "otk",
+        "pag",
+        "pal",
+        "pam",
+        "pan",
+        "pap",
+        "pau",
+        "pcd",
+        "pdc",
+        "phn",
+        "pli",
+        "pms",
+        "pol",
+        "por",
+        "ppl",
+        "prg",
+        "pus",
+        "quc",
+        "que",
+        "qxq",
+        "qya",
+        "rap",
+        "rif",
+        "roh",
+        "rom",
+        "ron",
+        "rue",
+        "run",
+        "rus",
+        "sag",
+        "sah",
+        "san",
+        "scn",
+        "sco",
+        "sgs",
+        "shs",
+        "shy",
+        "sin",
+        "sjn",
+        "slv",
+        "sma",
+        "sme",
+        "smo",
+        "sna",
+        "snd",
+        "som",
+        "spa",
+        "sqi",
+        "srd",
+        "srn",
+        "stq",
+        "sun",
+        "sux",
+        "swa",
+        "swe",
+        "swg",
+        "syc",
+        "syr",
+        "tah",
+        "tam",
+        "tat",
+        "tel",
+        "tet",
+        "tgk",
+        "tgl",
+        "tha",
+        "tig",
+        "tir",
+        "tkl",
+        "tlh",
+        "tly",
+        "tmh",
+        "tmr",
+        "toi",
+        "ton",
+        "tpi",
+        "tpw",
+        "tsn",
+        "tso",
+        "tts",
+        "tuk",
+        "tur",
+        "tvl",
+        "tyv",
+        "tzl",
+        "udm",
+        "uig",
+        "ukr",
+        "umb",
+        "urd",
+        "uzb",
+        "vec",
+        "vep",
+        "vie",
+        "vol",
+        "war",
+        "wln",
+        "wol",
+        "xal",
+        "xho",
+        "xmf",
+        "xqa",
+        "yid",
+        "yor",
+        "zea",
+        "zgh",
+        "zho",
+        "zul",
+        "zza"
+      ],
+      "source_url": "https://github.com/Helsinki-NLP/Tatoeba-Challenge/blob/master/data/README-v2021-08-07.md",
+      "asOf": "2026-06-28",
+      "confidence": "high",
+      "verified": "confirmed",
+      "publishedHeadline": 557,
+      "note": "336 ISO-639-3 codes enumerated from the v2021-08-07 pair table; the 557 headline counts script-varieties + the broader test-set inventory at finer granularity and isn't enumerable as distinct alpha-3 codes.",
+      "homepage": "https://github.com/Helsinki-NLP/Opus-MT",
+      "license": "CC-BY-4.0 (models, per the Helsinki-NLP releases)",
+      "nature": "Open per-pair research models (Helsinki-NLP, University of Helsinki) — hundreds of small downloadable models; a model for a given pair may not exist.",
+      "supersededBy": "hf-models → per-checkpoint Helsinki-NLP/opus-mt-* entries; family identity in shared/model-aliases.json"
+    },
+    {
+      "key": "tilde",
+      "label": "Tilde MT",
+      "count": 22,
+      "countBasis": "publisher-list-rows",
+      "iso6393Count": 22,
+      "countBasisNote": "Catalogued by machinetranslate.org rather than by Tilde directly — a secondary source, which is why confidence stays medium.",
+      "tier": "open",
+      "anyToAny": false,
+      "codeSystem": "ISO 639-1",
+      "iso6393": [
+        "ara",
+        "bul",
+        "ces",
+        "dan",
+        "deu",
+        "eng",
+        "est",
+        "fin",
+        "fra",
+        "hrv",
+        "hun",
+        "ita",
+        "lav",
+        "lit",
+        "nld",
+        "pol",
+        "por",
+        "ron",
+        "rus",
+        "spa",
+        "swe",
+        "ukr"
+      ],
+      "source_url": "https://machinetranslate.org/tilde",
+      "asOf": "2026-07-05",
+      "confidence": "medium",
+      "verified": "partially-confirmed",
+      "note": "General-purpose engine list as catalogued by machinetranslate.org; Tilde offers additional custom engines by request. API live-verified en→lv 2026-07-05.",
+      "homepage": "https://tilde.ai",
+      "nature": "Open research models (Tilde, EU) — per-pair systems for European and Baltic languages."
+    },
+    {
+      "key": "translated",
+      "label": "Translated (Lara)",
+      "count": 200,
+      "countBasis": "publisher-stated-headline",
+      "iso6393Count": null,
+      "countBasisNote": "Publisher says \"200+\"; we carry 200 as a floor with no enumerated list behind it. Nothing may be derived from this number.",
+      "tier": "service",
+      "anyToAny": true,
+      "codeSystem": "BCP-47 (base ISO 639-1 accepted)",
+      "iso6393": [],
+      "source_url": "https://developers.laratranslate.com/docs/supported-languages",
+      "asOf": "2026-07-05",
+      "confidence": "medium",
+      "verified": "publisher-count-only",
+      "note": "Publisher states 200+ languages; iso6393 list pending import (count cited) — same pending pattern as OPUS-MT.",
+      "homepage": "https://laratranslate.com",
+      "nature": "Commercial cloud API (Translated, Lara).",
+      "releaseDate": "2024-11-04",
+      "releaseDate_source_url": "https://translated.com/unveiling-lara"
+    },
+    {
+      "key": "m2m100",
+      "label": "M2M-100",
+      "count": 100,
+      "countBasis": "champollion-derived-enumeration",
+      "iso6393Count": 100,
+      "countBasisNote": "Our enumeration happens to equal Meta's own headline of 100 — coincidence of a clean list, not evidence that the two are the same kind of number.",
+      "tier": "open",
+      "anyToAny": true,
+      "codeSystem": "ISO 639-3",
+      "iso6393": [
+        "afr",
+        "amh",
+        "ara",
+        "ast",
+        "aze",
+        "bak",
+        "bel",
+        "ben",
+        "bos",
+        "bre",
+        "bul",
+        "cat",
+        "ceb",
+        "ces",
+        "cym",
+        "dan",
+        "deu",
+        "ell",
+        "eng",
+        "est",
+        "fas",
+        "fin",
+        "fra",
+        "fry",
+        "ful",
+        "gla",
+        "gle",
+        "glg",
+        "guj",
+        "hat",
+        "hau",
+        "heb",
+        "hin",
+        "hrv",
+        "hun",
+        "hye",
+        "ibo",
+        "ilo",
+        "ind",
+        "isl",
+        "ita",
+        "jav",
+        "jpn",
+        "kan",
+        "kat",
+        "kaz",
+        "khm",
+        "kor",
+        "lao",
+        "lav",
+        "lin",
+        "lit",
+        "ltz",
+        "lug",
+        "mal",
+        "mar",
+        "mkd",
+        "mlg",
+        "mon",
+        "msa",
+        "mya",
+        "nep",
+        "nld",
+        "nor",
+        "nso",
+        "oci",
+        "ori",
+        "pan",
+        "pol",
+        "por",
+        "pus",
+        "ron",
+        "rus",
+        "sin",
+        "slk",
+        "slv",
+        "snd",
+        "som",
+        "spa",
+        "sqi",
+        "srp",
+        "ssw",
+        "sun",
+        "swa",
+        "swe",
+        "tam",
+        "tgl",
+        "tha",
+        "tsn",
+        "tur",
+        "ukr",
+        "urd",
+        "uzb",
+        "vie",
+        "wol",
+        "xho",
+        "yid",
+        "yor",
+        "zho",
+        "zul"
+      ],
+      "source_url": "https://huggingface.co/facebook/m2m100_418M",
+      "asOf": "2026-07-19",
+      "confidence": "high",
+      "verified": "derived",
+      "publishedHeadline": 100,
+      "license": "MIT (checkpoints, per the model card)",
+      "homepage": "https://huggingface.co/facebook/m2m100_418M",
+      "nature": "Open research model (Meta) — direct many-to-many translation across its whole set, downloadable and self-hostable.",
+      "note": "Champollion-derived enumeration from the model card's published tag list (100 tags; script/region variants collapsed to the base language via the pinned LinguaMeta snapshot). The card's own headline is \"100 languages\". Snapshot: shared/catalogue/sources/.",
+      "releaseDate": "2020-10-21",
+      "releaseDate_source_url": "https://arxiv.org/abs/2010.11125",
+      "supersededBy": "hf-models → hf:facebook/m2m100_418M (owner’s card, declared extra — its pipeline tag hides it from the translation filter)"
+    },
+    {
+      "key": "madlad",
+      "label": "MADLAD-400",
+      "count": 415,
+      "countBasis": "champollion-derived-enumeration",
+      "iso6393Count": 415,
+      "countBasisNote": "count is OUR enumeration. Google's own headline is 419 (publishedHeadline) — never attribute 415 to Google.",
+      "tier": "open",
+      "anyToAny": true,
+      "codeSystem": "ISO 639-3",
+      "iso6393": [
+        "aar",
+        "abt",
+        "ace",
+        "acf",
+        "ach",
+        "ada",
+        "adh",
+        "ady",
+        "afr",
+        "agr",
+        "ahk",
+        "aka",
+        "akb",
+        "alt",
+        "alz",
+        "amh",
+        "amu",
+        "ang",
+        "ann",
+        "ape",
+        "ara",
+        "arn",
+        "asm",
+        "ava",
+        "awa",
+        "aym",
+        "aze",
+        "bak",
+        "bam",
+        "ban",
+        "bas",
+        "bbc",
+        "bci",
+        "bel",
+        "ben",
+        "bew",
+        "bgp",
+        "bgz",
+        "bho",
+        "bik",
+        "bim",
+        "bis",
+        "bjj",
+        "bod",
+        "bos",
+        "bqc",
+        "bre",
+        "bru",
+        "brx",
+        "bts",
+        "btx",
+        "bua",
+        "bul",
+        "bum",
+        "bus",
+        "bzj",
+        "cab",
+        "cac",
+        "cak",
+        "cat",
+        "cbk",
+        "cce",
+        "ceb",
+        "ces",
+        "cfm",
+        "cha",
+        "che",
+        "chk",
+        "chm",
+        "chr",
+        "chv",
+        "cjk",
+        "ckb",
+        "cnh",
+        "cni",
+        "cor",
+        "cos",
+        "crh",
+        "crs",
+        "ctu",
+        "cuk",
+        "cym",
+        "dan",
+        "deu",
+        "din",
+        "div",
+        "dje",
+        "djk",
+        "dov",
+        "dtp",
+        "dwr",
+        "dyu",
+        "dzo",
+        "ell",
+        "emp",
+        "eng",
+        "enq",
+        "epo",
+        "est",
+        "eus",
+        "ewe",
+        "fao",
+        "fas",
+        "ffm",
+        "fij",
+        "fil",
+        "fin",
+        "fip",
+        "fon",
+        "fra",
+        "frp",
+        "fry",
+        "ful",
+        "gag",
+        "gbm",
+        "gil",
+        "gla",
+        "gle",
+        "glg",
+        "glv",
+        "gof",
+        "gom",
+        "gor",
+        "grc",
+        "grn",
+        "gsw",
+        "gub",
+        "guc",
+        "guh",
+        "gui",
+        "guj",
+        "gvl",
+        "gym",
+        "gyn",
+        "hat",
+        "hau",
+        "haw",
+        "heb",
+        "hif",
+        "hil",
+        "hin",
+        "hmn",
+        "hmo",
+        "hne",
+        "hrv",
+        "hui",
+        "hun",
+        "hus",
+        "hvn",
+        "hye",
+        "iba",
+        "ibb",
+        "ibo",
+        "ify",
+        "iku",
+        "ilo",
+        "inb",
+        "ind",
+        "isl",
+        "iso",
+        "ita",
+        "ium",
+        "izz",
+        "jac",
+        "jam",
+        "jav",
+        "jax",
+        "jiv",
+        "jpn",
+        "jvn",
+        "kaa",
+        "kac",
+        "kal",
+        "kan",
+        "kas",
+        "kat",
+        "kaz",
+        "kbd",
+        "kbp",
+        "kek",
+        "kha",
+        "khm",
+        "kin",
+        "kir",
+        "kjg",
+        "kjh",
+        "kmb",
+        "knj",
+        "koi",
+        "kom",
+        "kon",
+        "kor",
+        "kos",
+        "krc",
+        "kri",
+        "krj",
+        "ksd",
+        "ksw",
+        "ktu",
+        "kua",
+        "kum",
+        "kur",
+        "kwi",
+        "laj",
+        "lao",
+        "lat",
+        "lav",
+        "lhu",
+        "lin",
+        "lit",
+        "lrc",
+        "ltg",
+        "ltz",
+        "lub",
+        "lug",
+        "lus",
+        "luz",
+        "mad",
+        "mag",
+        "mah",
+        "mai",
+        "mak",
+        "mal",
+        "mam",
+        "mar",
+        "mas",
+        "maz",
+        "mbt",
+        "mdf",
+        "mel",
+        "meo",
+        "meu",
+        "mey",
+        "mfe",
+        "mgh",
+        "min",
+        "miq",
+        "mkd",
+        "mkn",
+        "mlg",
+        "mlt",
+        "mni",
+        "mnw",
+        "mon",
+        "mps",
+        "mqy",
+        "mri",
+        "mrj",
+        "mrw",
+        "msa",
+        "msb",
+        "msi",
+        "msm",
+        "mtr",
+        "mya",
+        "myv",
+        "nav",
+        "nbl",
+        "nep",
+        "new",
+        "ngu",
+        "nhe",
+        "nia",
+        "nij",
+        "niq",
+        "nld",
+        "nnb",
+        "noa",
+        "noe",
+        "nog",
+        "nor",
+        "nso",
+        "nut",
+        "nya",
+        "nyu",
+        "nzi",
+        "oci",
+        "oji",
+        "ori",
+        "orm",
+        "oss",
+        "otq",
+        "pag",
+        "pam",
+        "pan",
+        "pap",
+        "pau",
+        "pck",
+        "pis",
+        "pol",
+        "pon",
+        "por",
+        "ppk",
+        "prk",
+        "pus",
+        "qub",
+        "quc",
+        "que",
+        "quf",
+        "quh",
+        "qup",
+        "quy",
+        "qvc",
+        "qvi",
+        "qvz",
+        "qxr",
+        "raj",
+        "rcf",
+        "rki",
+        "rmc",
+        "roh",
+        "rom",
+        "ron",
+        "run",
+        "rus",
+        "rwo",
+        "sag",
+        "sah",
+        "san",
+        "sda",
+        "seh",
+        "shn",
+        "shp",
+        "sin",
+        "sja",
+        "sjp",
+        "skr",
+        "slk",
+        "slv",
+        "sme",
+        "smo",
+        "smt",
+        "sna",
+        "snd",
+        "som",
+        "sot",
+        "spa",
+        "spp",
+        "sqi",
+        "srm",
+        "srn",
+        "srp",
+        "srr",
+        "ssw",
+        "stq",
+        "sun",
+        "sus",
+        "suz",
+        "swa",
+        "swe",
+        "sxn",
+        "syl",
+        "syr",
+        "tab",
+        "taj",
+        "tam",
+        "tat",
+        "tbz",
+        "tca",
+        "tcy",
+        "tdx",
+        "tel",
+        "teo",
+        "tet",
+        "tgk",
+        "tha",
+        "tir",
+        "tiv",
+        "tks",
+        "tlh",
+        "tll",
+        "toj",
+        "ton",
+        "trp",
+        "tsc",
+        "tsg",
+        "tsn",
+        "tso",
+        "tuc",
+        "tuk",
+        "tur",
+        "tvl",
+        "twu",
+        "tyv",
+        "tyz",
+        "tzh",
+        "tzj",
+        "tzo",
+        "ubu",
+        "udm",
+        "uig",
+        "ukr",
+        "urd",
+        "uzb",
+        "vec",
+        "ven",
+        "vie",
+        "wal",
+        "war",
+        "wln",
+        "wol",
+        "xal",
+        "xho",
+        "xmm",
+        "yap",
+        "yid",
+        "yor",
+        "yua",
+        "yue",
+        "zap",
+        "zha",
+        "zho",
+        "zne",
+        "zul",
+        "zza"
+      ],
+      "source_url": "https://huggingface.co/google/madlad400-3b-mt",
+      "asOf": "2026-07-19",
+      "confidence": "medium",
+      "verified": "derived",
+      "publishedHeadline": 419,
+      "license": "Apache-2.0 (per the model card)",
+      "homepage": "https://huggingface.co/google/madlad400-3b-mt",
+      "nature": "Open research model (Google) — T5-based MT trained on 400+ languages; downloadable and self-hostable. Tail-language quality varies widely.",
+      "note": "Champollion-derived enumeration from the model card's published tag list (418 tags; script/region variants collapsed to the base language via the pinned LinguaMeta snapshot). The card's own headline is \"419 languages\". 2 published tag(s) had no LinguaMeta ISO 639-3 mapping and are not counted: zxx, ber. Snapshot: shared/catalogue/sources/.",
+      "releaseDate": "2023-09-09",
+      "releaseDate_source_url": "https://arxiv.org/abs/2309.04662",
+      "supersededBy": "hf-models → hf:google/madlad400-7b-mt (owner’s card, 416 languages vs 415 here)"
+    },
+    {
+      "key": "omt1600",
+      "label": "Omnilingual MT",
+      "count": 1600,
+      "countBasis": "publisher-stated-headline",
+      "iso6393Count": null,
+      "countBasisNote": "1,600 is the paper's own headline for languages it translates FROM, with no enumerable list published. Two further figures from the same abstract must never be collapsed into it: the models \"understand sufficiently well\" just over 400 (see deployable), and translate INTO about 1,200. Our ~1,200 long-tail figure is arithmetic (1,600 − 400+) and is a DIFFERENT ~1,200 from the paper's into-1,200 — always carry the arithmetic inline.",
+      "tier": "open",
+      "anyToAny": false,
+      "codeSystem": "languoids (paper)",
+      "iso6393": [],
+      "verified": "publisher-count-only",
+      "note": "iso6393 is empty because no per-language enumeration of the 1,600 is published — there is nothing upstream to ingest. The paper's only per-language list is Table D.1 (Appendix D), whose own caption scopes it to the languages covered by its evaluation/training artifact contributions, not to the 1,600. Count is cited from the publisher; this stays a count-only entry unless Meta publishes the full list.",
+      "source_url": "https://arxiv.org/abs/2603.16309",
+      "asOf": "2026-07-22",
+      "homepage": "https://ai.meta.com/research/publications/omnilingual-mt-machine-translation-for-1600-languages/",
+      "license": null,
+      "nature": "Research breadth model (The Omnilingual MT Team, FAIR at Meta; arXiv:2603.16309, 2026-03-17). Claims translation FROM 1,600 languages and into ~1,200 with non-trivial performance; no public weight release as of 2026-07. COUNT-ONLY entry: no per-language list ingested, so it never lights map coverage or joins the tier counts — it exists for the cited omnimodel-escalation timeline. Distinct from Omnilingual ASR (arXiv:2511.09690, 2025-11).",
+      "releaseDate": "2026-03-17",
+      "releaseDate_source_url": "https://arxiv.org/abs/2603.16309",
+      "deployable": {
+        "count": 400,
+        "note": "Meta’s own report: the models “understand sufficiently well” just over 400 of the 1,600 languages (§1); “non-trivial performance” translating into ~1,200. Their words — the paper never claims 1,600 deployable-quality languages.",
+        "source_url": "https://arxiv.org/abs/2603.16309"
+      }
+    }
+  ]
+}