ModelCards

Qwen3-Embedding-8B

Embedding modelOpen weightsCard summarized

Specifically designed for text embedding, retrieval, and reranking tasks

Source: Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models (technical report, arXiv:2506.05176) and Hugging Face model card

Use it via API

Self-host only: no published API price.
Code exampleCompare

Where to call it

  • Self-host (weights)Qwen/Qwen3-Embedding-8B
  • Alibaba Cloud Model Studiotext-embedding-v4

Supports

No feature list in the maker's docs.

Full feature support

Good for

  • Text retrieval and RAG
  • Code retrieval
  • Multilingual and cross-lingual retrieval

Watch out for

Overview

Alibaba · 01Embedding, front

Qwen3 Embed 8B

Specifically designed for text embedding, retrieval, and reranking tasks

Typical price
$0.01 / 1M tok
Cheapest host: DeepInfra
Dimensions
4096
Max input
32K tokens
Modes
text embeddings, code retrieval
Released
5 Jun 2025

Qwen3-Embedding-8B is the largest model in Alibaba's open Qwen3 Embedding series (0.6B, 4B, 8B, plus matching rerankers), built on the Qwen3-8B base model and released under Apache 2.0. It has a 32K context, outputs up to 4096 dimensions (any size from 32 up), accepts task instructions, and covers 100+ natural and programming languages. At release it ranked No. 1 on the MTEB multilingual leaderboard with a score of 70.58. model card; technical report

Qwen/Qwen3-Embedding-8B
Typical price
$0.01 / 1M tok via DeepInfra
Dimensions
4096 (can be set anywhere from 32 to 4096)
Max input
32K tokens
Languages
100+ languages (including programming languages)
Weights
Open · Apache 2.0
Released
5 Jun 2025
  • text embeddings
  • code retrieval
  • cross-lingual retrieval