“Find every named language model released since November 2022, and for each one give me: model name, the organization that made it, release date, its exact license, whether the weights are publicly downloadable, parameter count, and any restriction on commercial use. Include small and lesser-known models, not just the major labs. One row per model.”
A live-web census of 1,758 distinct named language-model records with release dates from November 2022 through September 2026 — major labs, small teams, regional and specialist releases alike. 1,056 have publicly downloadable weights, 238 do not, and 464 are unresolved; blank cells mean unknown, not “not applicable.” This is best-effort research, not legal advice.
“Exact license” is the model-weight license or named proprietary/custom terms where a source states one; public weight download is independent of whether that license is OSI-approved. “Commercial restriction” records model-specific limits — normal license obligations still apply even when this cell is blank. Only 282 of 1,758 records carry an explicitly named license and 295 a parameter count; the rest are left blank rather than guessed. Records dated 2026 often rest on sparse or secondary sourcing — verify through the linked source before legal or production decisions. Click a header to sort.
| Model | Organization | Released | Exact license | Weights | Parameters | Commercial restriction | Source |
|---|
Method: best-effort live-web census of named language-model records, 1,758 rows, release dates 2022-11 through 2026-09-03, one row per normalized model name. Release dates keep source precision (year, month, or day); a record's source link is the best-quality page among those mentioning it. The model ecosystem has no authoritative registry; unresolved license, weight, parameter, and restriction values are shown blank rather than invented. Compiled September 2026. Research, not legal advice.