Title: Mistoqsijiet frequenti %E2%80%94 LLM EU
Canonical: https://llmeu.com/mt/learn/
Markdown mirror: https://llmeu.com/mt/learn.md
Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json

# Mistoqsijiet frequenti

## Ba%C5%BCi%C4%8Bi

### [X'inhi tieqa tal-kuntest, u x'inhu token?](/mt/learn/context-window-and-tokens/)

Token huwa bi%C4%8B%C4%8Ba %C5%BCg%C4%A7ira ta' test, bejn wie%C4%A7ed u ie%C4%A7or framment ta' kelma. Tieqa tal-kuntest hija n-numru massimu ta' tokens li sej%C4%A7a wa%C4%A7da tista' %C5%BC%C5%BComm: istruzzjonijiet, il-konver%C5%BCazzjoni s'issa, passa%C4%A1%C4%A1i rkuprati u r-risposta.

### [LLM vs AI: x'inhu mudell lingwistiku kbir, x'inhu ba%C5%BCi llm, u x'inhu Nemotron 3 Super?](/mt/learn/llm-vs-ai/)

L-AI hija l-qasam; mudell lingwistiku kbir huwa tip wie%C4%A7ed ta' mudell fiha, LLM ba%C5%BCi jkompli t-test minflok ma jsegwi l-istruzzjonijiet, u Nemotron 3 Super huwa ta%C4%A7lita ta' esperti ta' 123.6B.

### [X'inhu LLM?](/mt/learn/what-is-an-llm/)

LLM huwa programm im%C4%A7arre%C4%A1 fuq ammonti kbar ta' test biex ibassar it-token li jmiss, bi%C4%8B%C4%8Ba %C5%BCg%C4%A7ira ta' test. Huwa jwie%C4%A1eb mill-pi%C5%BCijiet tieg%C4%A7u u l-prompt li tibg%C4%A7at, mhux minn database li jfittex fiha.

## Retrieval

### [X'inhu LLM retriever, u liema mudell ta' embedding multilingwi g%C4%A7andi nu%C5%BCa?](/mt/learn/embeddings-and-vector-search/)

Retriever jg%C4%A7aqqad mudell ta' embedding li jittrasforma t-test f'vectors ma' reranker li jag%C4%A7ti punte%C4%A1%C4%A1 lill-passa%C4%A1%C4%A1i. multilingual-e5-large huwa MIT b'chunks ta' 512-token; embeddinggemma-300m jie%C4%A7u 2048.

## Hosting

### [Kif ja%C4%A7dmu l-LLMs fl-UE?](/mt/learn/how-llms-run-in-the-eu/)

Mudell ja%C4%A7dem fejn ja%C4%A7dem endpoint, u r-ringiela tirre%C4%A1istra r-re%C4%A1jun u l-klassi tieg%C4%A7u. Ir-residenza hija filtru applikat qabel il-punte%C4%A1%C4%A1: jekk xejn ma jissodisfaha, it-talba tirritorna 409 no_endpoint_for_policy.

### [Open weights jew API: liema g%C4%A7andi nu%C5%BCa?](/mt/learn/open-weights-vs-api/)

Open weights tfisser li l-fajls jistg%C4%A7u jitni%C5%BC%C5%BClu; endpoint tfisser li xi %C4%A7add imexxihom g%C4%A7alik. %C4%A6afna skjeramenti Ewropej ju%C5%BCaw it-tnejn: API g%C4%A7al tag%C4%A7bija varjabbli, checkpoint self-hosted g%C4%A7al xog%C4%A7ol sensittiv.

## Spi%C5%BCa

### [X'jiswa sej%C4%A7a tal-API tal-LLM, u xi jfissru n-numri tal-benchmark?](/mt/learn/llm-cost-and-pricing/)

Prezz jinqara mir-rata li tirre%C4%A1istra ringiela ta' endpoint, g%C4%A7al kull miljun token 'il %C4%A1ewwa u 'l barra, flimkien ma' tariffa wa%C4%A7da vi%C5%BCibbli g%C4%A7al kull talba. Ringiela ming%C4%A7ajr rata rre%C4%A1istrata tintwera ming%C4%A7ajr prezz.

## Li%C4%8Benzji

### [Kif ja%C4%A7dmu l-li%C4%8Benzji tal-mudelli miftu%C4%A7a?](/mt/learn/llm-licences-explained/)

Li%C4%8Benzja hija qasam irre%C4%A1istrat fuq kull karta tal-mudell: identifikatur SPDX u link g%C4%A7at-test. Pi%C5%BCijiet miftu%C4%A7a jfisser li l-pi%C5%BCijiet jistg%C4%A7u jitni%C5%BC%C5%BClu, mhux li huwa permess u%C5%BCu kummer%C4%8Bjali.

## API

### [X'inhuma l-limiti tar-rata tal-API tal-LLM?](/mt/learn/llm-rate-limits-explained/)

Limitu tar-rata hawnhekk huwa limitu massimu g%C4%A7al kull barmil f'tieqa fissa: 600 sej%C4%A7a g%C4%A7al kull %C4%8Bavetta u 1800 g%C4%A7al kull organizzazzjoni kull minuta, b'massimu ta' 8 talbiet pendenti fuq %C4%8Bavetta wa%C4%A7da.

## Routing

### [Kif ja%C4%A7dem router tal-LLM?](/mt/learn/llm-routing-explained/)

Router jibdel talba li ma ssemmi l-ebda magna f'rispons ta' "endpoint" wie%C4%A7ed. Funzjoni pura tiffiltra l-kandidati skont il-politika u r-residenza, tiskorja dak li jibqa' %C4%A7aj u tie%C4%A7u l-og%C4%A7la skor.

## Irkupru

### [X'inhu RAG?](/mt/learn/what-is-rag/)

RAG hija pipeline, mhux mudell: l-applikazzjoni tieg%C4%A7ek tirkupra siltiet u l-mudell iwie%C4%A1eb minn dawk. L-a%C4%A7%C4%A7ar stadju biss huwa tlestija ta%C4%8B-chat; l-inkorporazzjoni u r-reranking huma sej%C4%A7iet separati.