Report · 8 October 2026
State of Kashmiri AI, 2026
Where Kashmiri stands in AI and language technology this year, measured from the 205 resources in our open catalogue and from checks of the tools people use every day.
Key findings
- 205 Kashmiri datasets, models, papers, tools and research groups are catalogued as of 8 October 2026.
- The field is young: 101 of the 173 dated entries (58%) appeared in 2025 or 2026. In 2022 there were 9; 2026 has 69 so far.
- Work is concentrated: Machine Translation (73), Corpus (49), LLM (35) lead, while 5 tasks have five entries or fewer.
- 119 entries (58%) live on Hugging Face; 60 (29%) state no licence at all.
- Support in everyday tools is uneven: Google Translate, Windows, Whisper (OpenAI) do not support Kashmiri, while phone keyboards and Microsoft Translator do.
A field that is just starting
Resources by year of publication. The catalogue holds only 8 dated entries from before 2022; since then the count has risen almost every year. 24 entries have no publication year and are left out of the chart, as are research groups, whose year is a founding date.
What exists
- 205
- Resources
- 84
- Datasets
- 60
- Models
- 36
- Papers
By task
An entry can serve several tasks, so these add up to more than 205. Tasks with a page of their own link to it.
Where almost nothing exists
5 tasks have five catalogued entries or fewer. These are the clearest openings for new work on Kashmiri.
- 5Survey
- 5Evaluation
- 5Diacritization
- 1Sentiment Analysis
- 1Text Input
Licences
88 entries (43%) carry an open licence such as Apache-2.0, MIT or CC BY. 60 (29%) state none, which in practice means you must ask before reusing them. “See source” covers entries whose terms are set by a publisher or a dataset card.
Kashmiri in everyday tools
Checked for our guides in October 2026. Each row links to the write-up and its sources.
| Tool | Kashmiri? | Detail |
|---|---|---|
| Google Translate | No | Kashmiri is not among its languages. Source |
| Microsoft Translator | Yes | Lists Kashmiri (ks) for text translation. Source |
| NLLB-200 (Meta) | Yes | Open model; covers Kashmiri in Perso-Arabic and Devanagari. Source |
| IndicTrans2 (AI4Bharat) | Yes | Open model; supports both scripts. Source |
| Gboard on Android | Yes | Kashmiri keyboards in Arabic and Devanagari script, with transliteration. Source |
| iPhone, iPad and Mac keyboards | Yes | Kashmiri (Arabic) and Kashmiri (Devanagari) layouts. Source |
| Windows | No | No Kashmiri layout; adding Kashmiri gives the Urdu keyboard. Source |
| Whisper (OpenAI) | No | Kashmiri is not in its speech-recognition language list. Source |
| MMS (Meta) | Partial | Language identification only; no speech recognition or synthesis. Source |
| Mozilla Common Voice | Partial | Listed, but not yet open for recording. Source |
| Unicode | Yes | Kashmiri yeh fixed in 16.0; Sharada Kashmiri vowel signs added in 17.0 (2025). Source |
Method
Every figure above is computed from the Kashmiri NLP Resource Hub as it stood on 8 October 2026. The hub lists what we and contributors could find; it is not a census, and work that is unpublished or not online is missing. Counts include an entry once per kind and once per task it serves. To add or correct an entry, submit it.
Cite this report
Kashmir AI Research (2026). State of Kashmiri AI, 2026. 8 October 2026. https://kashmirairesearch.online/state-of-kashmiri-ai-2026