Qwen3-Embedding-8B

Alibaba's Apache-2.0 open-weights multilingual text embedding models in 0.6B, 4B and 8B sizes.

Where it ranks

Vendor: Alibaba (Qwen)
Family: Qwen3-Embedding
Released: Jun 2025
Weights: Open: download and run it yourself
License: Apache-2.0
Parameters: 7.57B
GPU memory to run locally: ~16 GB (Derived: 7.57B parameters x 2 bytes = ~15.1 GB of bf16 weights; 4B and 0.6B sizes and GGUF builds also exist.)
Access: Free (open weights)
Last verified: 19 Oct 2026

Benchmark	Result	Entry on the leaderboard	As of	Source
Agentset Embedding Model Leaderboard	1,510	`Qwen3 Embedding 8B` ELO from GPT-5-judged pairwise comparisons; rank 8 of 18; win rate 48.8%; page says 'Last updated: February 15, 2026'.	15 Feb 2026	agentset.ai
MTEB(eng, v2)	75.2%	`Qwen/Qwen3-Embedding-8B` Mean (Task) over all 41 tasks; board rank 5 (Borda); board zero-shot 95%; read from the leaderboard's public API.	19 Oct 2026	mteb-leaderboard-backend.hf.space
MTEB(Multilingual, v2)	70.6%	`Qwen/Qwen3-Embedding-8B` Mean (Task) over all 131 tasks; board rank 4 (Borda); board zero-shot 99%; read from the leaderboard's public API.	19 Oct 2026	mteb-leaderboard-backend.hf.space