mxbai-embed-large-v1
Free.ai (self-hosted)
·
embeddings
·
~100 tokens per call
mxbai-embed-large-v1 ke an embedding model e hahiloeng ke mixedbread.ai. E ka ba e le Semantic search, clustering, similarity.. E na le li-Free.ai GPUs tse nang le li-server tse 100 tse nang le li-server tse 100 tse nang le li-server tse 100. E lokollotsoe ka Apache 2.0 — ho sebelisoa ka khoebo ho lumelloa ka Free.ai.
Ho sebelisa ka API
REST API e lumellanang le OpenAI. E etsa konopo'me e bitsa mofuta ona ka metsotsoana.
curl -X POST https://api.free.ai/v1/image/generate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"mxbai-embed-large-v1","prompt":"your prompt here"}'Lipotso tse botsoang khafetsa
mxbai-embed-large-v1 e fetolela mongolo ho becquerel e thata (lista ea li-floats) e fang moelelo. E sebelise bakeng sa ho batla ka semantic, ho kopanya, ho khothaletsa, ho fumana ho eketseha ho theha (RAG), le mesebetsi efe kapa efe moo "na mongolo ona o tšoana le mongolo oo" e leng bohlokoa.
Li-dimensions tse tloaelehileng ke 384, 768, 1024, kapa 1536 ho latela mofuta. BGE-M3 e fana ka 1024-dim; OpenAI Ada e fana ka 1536. Likarabo tsa API li kenyelletsa boholo ka hona vector DB ea hau e khetha index e nepahetseng.
Litšoantšo tsa embedding tsa morao-rao (ho akarelletsa le likhetho tse ngata ho Free.ai) li koetlisitsoe ka lipuo tse fetang 100. Likarolo tsa ho fumana lipuo tse fapaneng - ho batla ka Senyesemane, ho hokahanya litokomane ka Sepanishe.
512 ho 8,192 tokens ho itšetlehile ka mofuta. Inputs khutšoanyane ke truncated — chunk litokomane tse telele ho liparagrapa pele embedding.
mxbai-embed-large-v1 e sebetsa ka GPUs ea rona le ke e 'ngoe ea lisebelisoa tse theko e tlaase - ka li-token tse ka bang 100 ka ho ngola tse tsoang ho pool ea hau ea mahala ka letsatsi. $ 5 = li-token tsa 200K.
E-na — POST lethathamo la lihlooho ho /v1/embeddings/ le mxbai-embed-large-v1 e khutlisa lethathamo la li-vectors ka tsela e ts'oanang. Boholo ba batch ho fihlela ho 2,048 ka kopo.
L2-normalized ka ho tloaeleha — cosine similarity = dot product. Lekola `normalize=false` haeba u batla li-vectors tse sa phekoloeng bakeng sa li-metrics tse fapaneng tsa bophahamo.
Any — Pinecone, Weaviate, Qdrant, Chroma, pgvector, FAISS, LanceDB. mxbai-embed-large-v1 returns plain JSON floats; the DB never sees the model.
E-na — POST ho /v1/embeddings/ le model="mxbai-embed-large-v1". OpenAI-compatible response shape, ka hona libuka tsa bareki tse teng li sebetsa ka nepo. /api/ e na le litlhaloso tse felletseng.
Li-model tse sebete li boloka mongolo oa hau ho GPUs tsa rona'me li o tlohela ha ho khutla ho ngola. Premium e tsamaea ka DPA. Re sa koetlise ka li-input tsa hau.
Sub-100ms bakeng sa mongolo o khuts'oane ka ho arolelana, 100-500ms ka premium. Li-call tsa batch li theoha ka mokhoa o hlakileng - li-chunks tsa 1,000 li felile ka metsotsoana ea 2-10.
Ee - Free.ai e fana ka ts'ebeliso ea khoebo ea embeddings. E theha ho batla tlhahiso, li-pipelines tsa RAG, li-systems tsa litaelo ntle le royalty ea per-vector.