Qwen3-Embedding-8B

7.57B parameters, text in, embeddings out, 40K tokens of context. 2 providers serve it from $0.010 per million input tokens.

Verified 2026-10-25 Rank #11 in Embeddings · score 104.7 method v1 Sources 3 Downloads 30d 2M · #63 of 619 tracked

Changed this week: 3 new quantized builds on the Hub (44 in all).2026-10-24 · 2 new quantized builds on the Hub (43 in all).2026-10-24

Can I run it?4.5GB at INT4Estimatefits a 16 GB cardmethod

What does it cost?$0.010/1M incheapest of 1DeepInfra2026-10-25

Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-24

Does it work?51.56on MTEB Reranking#1 of 33 measuredleaderboard2026-10-24

Do people use it?2Mdownloads, 30 days#63 of 619 we track · top 10%+6 likes, downloads −4% this week840 likesHugging Face2026-10-24

Worth reading

Picked by rule from 9 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Latest · 35 days ago

How Qwen3-Embedding-8B Reached #1: The Evolution of Text Embeddings - DEV Community

Qwen3-Embedding-8B reached #1 on the MTEB multilingual leaderboard in June 2025 with a score of 70.58.

The embedding model uses the hidden representation corresponding to the final EOS token to produce the embedding.
Qwen3 provides both embedding and reranking models, including 0.6B, 4B and 8B variants.

dev.to 2026-09-21 2.2k words

9 best of 9 tracked

The 1 pick above, plus these 8. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleQwen3 Embedding:Advancing Text Embedding and Reranking Through Foundation Modelsqwenlm.github.io · 2025-06-05 · 3 HN pts · official
PaperQwen3 Embedding: Advancing Text Embedding and Reranking with Foundation Modelsarxiv.org · 2025-06-11 · 1 HN pts
ArticleFine-Tuning Qwen3 Embeddings for product category classificationblog.ivan.digital · 2026-03-03 · 2 HN pts
ArticleQwen3-Embedding-8B Retrieveremergentmind.com · undated
ArticleQwen3-Embedding-8B: Multilingual Text Embeddingsemergentmind.com · undated
ArticleQwen3-Embedding-8B/4096 Model Overviewemergentmind.com · undated
CodeQwenLM/Qwen3-Embeddinggithub.com · 2025-06-05
ArticleAlibaba Qwen Team Releases Qwen3-Embedding and Qwen3-Reranker Series – Redefining Multilingual Embedding and Ranking Standards - MarkTechPostmarktechpost.com · 2025-06-06

Show all 8

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

#	Cheapest 2	In /1M	Out /1M	Context	Quant	Seen
1	DeepInfra	$0.010	—	32K	—	Oct 25
2	Scaleway	—	—	—	—	Oct 25

All 2 providers →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Served by 2 API providers; 44 quantized builds for local use; a big base for builders (63 fine-tunes, adapters and merges).

As a service2DeepInfra, Scaleway
On own hardware44quantized builds
Built on6346 fine-tunes · 17 adapters · 0 merges
In apps62Hugging Face Spaces

Top Spaces running it: embedl/hfviewer ♥ 181 · emanforhelp/model-pulse ♥ 5 · raidium/jolia ♥ 4 · alx-d/PhiRAG ♥ 2

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 16 GB card at INT4

About 4.5 GB at INT4, 9.1 GB at FP8, 18 GB at BF16: weights only, so leave room for the KV cache.

Estimate params × bytes per param × 1.2 · how we estimate

How it scores

Third-party measurements only; author-reported numbers stay in Reference, marked as such.

#1 of 33 on MTEB Reranking so far

51.56. Each tick is a model on this site with a measured MTEB Reranking row; author-reported numbers stay in Reference. MAP x100 (mean of 2 tasks). 2 tasks, split=test

Source · leaderboard · 2026-10-24

The story so far

Weights, every dated article, and repo commits on one rail.

Most coverage came in the first two weeks

Teal is the weights (2025-06-03), amber triangles are 6 dated articles (each links to it), grey ticks are repo commits.

What people built on it

107 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

107 derivatives on the Hub, led by 46 fine-tunes

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF mradermacher/Qwen3-Embedding-8B-i1-GGUF for llama.cpp, Ollama, LM Studio 8.4k downloads MLX mlx-community/Qwen3-Embedding-8B-4bit-DWQ for Apple silicon 47.1k downloads AWQ drawais/Qwen3-Embedding-8B-AWQ-INT4 for vLLM, SGLang 213.7k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

4 rows

Benchmarks 4 rows

measured third-party leaderboard row (3)author-reported number from the model card or eval files (1)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
measured MTEB · Embed text for search	69.44 nDCG@10 x100 (mean of 10 tasks)	10 tasks, split=test	leaderboard	2026-10-24
measured MTEB · Embed text for search	75.22 mean(task) x100 over 41 tasks	41 tasks, split=test	leaderboard	2026-10-24
measured MTEB Reranking · Rerank search results	51.56 MAP x100 (mean of 2 tasks)	2 tasks, split=test	leaderboard	2026-10-24
author-reported MTEB · Embed text for search	70.58	8B size	model author	2026-10-07

What the card says

1 cited sentence

Summary generated, cited

The model is designed for tasks such as text retrieval and classification, supporting over 100 languages.^[1]^[2]

“text retrieval, code retrieval, text classification, text…”
“Supported Languages: 100+ Languages”

Generated 2026-10-07 15:00 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

5 use cases

What it is good at 5 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Embed text for searchmeasuredMTEB(eng, v2) retrieval 69.44 MTEB(eng, v2) 75.22 Author reports MTEB multilingual leaderboard 70.58 +5 more
Call tools and APIsstructuralchat template: <tool_call>
Chat and follow instructionsstructuralchat template present
Run on a laptop or phonestructuralfits: single-gpu
Work across languagesstructuraldecision-model estimate p=1.00

Facts

20 fields

Listed in Embeddings

Facts

Qwen3-Embedding-8B has 7.57B parameters. (safetensors metadata, observed 2026-10-24)
Context length is 40,960 tokens (40K). (config.json max_position_embeddings, observed 2026-10-24)
License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
Created 2025-06-03, last modified 2025-07-07. (HF API, observed 2026-10-24)

Pipeline tag	feature-extraction
Library	sentence-transformers
Architectures	Qwen3ForCausalLM
Model type	qwen3
Dtype	bfloat16
Params by dtype	BF16: 7,567,295,488
Size band	M
MoE	no
Context length	40,960 (config.json max_position_embeddings)
Papers	arXiv:2506.05176
License	Apache 2.0 (apache-2.0)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen3ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for Qwen3-Embedding-8B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	18 GB · Single GPU (<24 GB)
FP8 (1 byte/param)	9.1 GB · Single GPU (<24 GB)
INT4 (0.5 bytes/param)	4.5 GB · Single GPU (<24 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

40 variants

Lineage

Fine-tuned from Qwen/Qwen3-8B-Base (same organization) Family: Qwen/Qwen3-Embedding 107 derivatives on the Hub

Base model: Qwen/Qwen3-8B-Base

finetune: 46quantized: 44adapter: 17

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
Octen/Octen-Embedding-8B	finetune	—	7.57B	M	apache-2.0	Octen	194	1.5M
farbodtavakkoli/OTel-Embedding-8B	finetune	—	7.57B	M	apache-2.0	farbodtavakkoli	1	264.1k
drawais/Qwen3-Embedding-8B-AWQ-INT4	quantized	AWQ	8.24B	M	apache-2.0	drawais	5	213.7k
alexliap/Qwen3-Embedding-8B-NVFP4	quantized	NVFP4	4.53B	M	apache-2.0	alexliap	8	97.8k
Maxi-Lein/Qwen3-Embedding-8B-onnx	quantized	—	—	—	apache-2.0	Maxi-Lein	2	69.6k
mlx-community/Qwen3-Embedding-8B-4bit-DWQ	quantized	-4bit	7.57B	M	apache-2.0	mlx-community	11	47.1k
alexliap/Qwen3-Embedding-8B-FP8-DYNAMIC	quantized	-FP8	7.57B	M	apache-2.0	alexliap	3	22k
boboliu/Qwen3-Embedding-8B-W4A16-G128	quantized	-W4A16	8.24B	M	apache-2.0	boboliu	4	17.9k
groxaxo/qwen3-embed-8b-gptq	quantized	gptq	7.57B	M	apache-2.0	groxaxo	2	16.4k
furiosa-ai/Qwen3-Embedding-8B	finetune	—	7.57B	M	apache-2.0	furiosa-ai	0	10.3k
mradermacher/Qwen3-Embedding-8B-i1-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	3	8.4k
huynguyendbs/Qwen3-Embedding-8B-4bit-MLX	quantized	-4bit	8.19B	M	apache-2.0	huynguyendbs	4	7.5k
mradermacher/Qwen3-Embedding-8B-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	2	7.2k
Octen/Octen-Embedding-8B-INT8	quantized	INT8	7.57B	M	apache-2.0	Octen	7	6.9k
zenlm/zen-embedding-8B-GGUF	quantized	GGUF	—	—	apache-2.0	zenlm	0	6.5k
Mira190/Euler-Legal-Embedding-V1	finetune	—	8.19B	M	apache-2.0	Mira190	17	5.2k
majentik/Qwen3-Embedding-8B-GGUF-Q4_K_M	quantized	GGUF	—	—	apache-2.0	majentik	0	5.8k
ZanderM98/Qwen3-Embedding-8B-bnb-4bit	quantized	-bnb	7.78B	M	apache-2.0	ZanderM98	2	5.6k
jedisct1/Qwen3-Embedding-8B-mlx	quantized	mlx	8.19B	M	apache-2.0	jedisct1	2	4.9k
cstr/qwen3-embed-8b-GGUF	quantized	GGUF	—	—	apache-2.0	cstr	0	4.5k
sionic-ai/comsat-embed-ja-8b-preview	finetune	—	7.57B	M	cc-by-nc-4.0	sionic-ai	2	2.6k
enacimie/Qwen3-Embedding-8B-Q4_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	enacimie	0	2.3k
LingoIITGN/qwen-indic-v1	adapter	—	7.57B	M	apache-2.0	LingoIITGN	3	1.9k
sionic-ai/comsat-embed-ko-8b-preview	finetune	—	7.57B	M	cc-by-nc-4.0	sionic-ai	19	1k
gabor-hosu/Qwen3-Embedding-8B-bnb-4bit	quantized	-bnb	7.78B	M	apache-2.0	gabor-hosu	0	1.9k
LostGentoo/Qwen3-Embedding-8B-AWQ	quantized	AWQ	7.62B	M	apache-2.0	LostGentoo	1	1.8k
ThakiCloud/SKILLRET-Embedding-8B	finetune	—	7.57B	M	apache-2.0	ThakiCloud	4	1.2k
endyjasmi/Qwen3-Embedding-8B-Q4_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	endyjasmi	1	1.3k
batiai/Qwen3-Embedding-8B-GGUF	quantized	GGUF	—	—	apache-2.0	batiai	0	1.3k
QuaduxIT/Qwen3-Embedding-8B-W8A16	quantized	—	7.57B	M	apache-2.0	QuaduxIT	0	1.2k
uw-math-ai/MathLeap-Qwen-8B	finetune	—	7.57B	M	apache-2.0	uw-math-ai	1	959
bsisduck/Qwen3-Embedding-8B-fp16-mlx	quantized	-fp16	7.57B	M	apache-2.0	bsisduck	0	1k
Chiwei/Qwen3-Embedding-8B-Q4_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	Chiwei	0	940
aninokumar/Qwen3-8B-M2V-Entropy-RAG	finetune	—	621M	S	apache-2.0	aninokumar	2	724
amkdg/Qwen3-Embedding-8B-NVFP4	quantized	NVFP4	5.15B	M	apache-2.0	amkdg	0	803
SultanGurbuz/qwen3-embedding-8b-finetuned	quantized	GGUF	7.58B	M	—	SultanGurbuz	0	795
majentik/Qwen3-Embedding-8B-GGUF-Q8_0	quantized	GGUF	—	—	apache-2.0	majentik	0	749
majentik/Qwen3-Embedding-8B-MLX-8bit	finetune	MLX	—	—	apache-2.0	majentik	0	720
TiGa-RCE/Qwen3-Embedding-8B-MLX-BF16	finetune	MLX	7.57B	M	apache-2.0	TiGa-RCE	1	649
hugger7326/Qwen3-Embedding-8B-Q8_0-GGUF	quantized	Q8_0	—	—	apache-2.0	hugger7326	0	650

Repository changelog

15 commits

changed in the last 7 daysOther: 2026-10-24, 2026-10-24all changes

Repository changelog 15 commits

Date	Commit	SHA
2025-07-07	Create LICENSE	1d8ad4c
2025-06-20	Update README.md with TEI support (#16)	80946ea
2025-06-07	add use case for vllm & modify Citation (#5)	a3d38e3
2025-06-06	Automatically add EOS via Tokenizer, integrate Sentence Transformers (#1)	4e42393
2025-06-05	Update README.md	22e872b
2025-06-05	Update README.md	14492ee
2025-06-05	Delete special_tokens_map.json	d3fb60a
2025-06-05	Delete added_tokens.json	6e05bb1
2025-06-05	Update README.md	ba49875
2025-06-05	Update README.md	c90816d
2025-06-05	Update README.md	9782d53
2025-06-05	Update README.md	6feafac
2025-06-05	Update config.json	263cad8
2025-06-03	Upload folder using huggingface_hub	d74263f
2025-06-03	initial commit	41675d7

All providers

2 providers · 2 endpoints

Where to run 2 providers · 2 endpoints

Hosted endpoints observed for Qwen3-Embedding-8B, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.01/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
DeepInfra hf, deepinfra	Qwen/Qwen3-Embedding-8B	32K	—	$0.01	—	—	—	—	live	2026-10-25
Scaleway hf	qwen3-embedding-8b	—	—	—	—	—	—	—	live	2026-10-25

Articles and coverage

9 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official1

Qwen3 Embedding:Advancing Text Embedding and Reranking Through Foundation Models

qwenlm.github.io2025-06-05officialauthorityabout the Qwen3 Embedding familyHN 3 pts · 0 comments937 words

The Qwen3-Embedding-8B model ranks No.1 in the MTEB multilingual leaderboard with a score of 70.58. The Qwen3-Reranker-4B model achieves a FollowIR score of 14.84.generated, cited

Analysis8

How Qwen3-Embedding-8B Reached #1: The Evolution of Text Embeddings - DEV Community

dev.to2026-09-212,177 words

Qwen3-Embedding-8B reached #1 on the MTEB multilingual leaderboard in June 2025 with a score of 70.58. The model employs a dual-encoder architecture and a three-stage training strategy that leverages the Qwen3 foundation to generate synthetic training pairs. It features instruction awareness, supporting task-specific vectors, and includes Matryoshka representations alongside a 32K-token context window.written, reviewed, cited

The embedding model uses the hidden representation corresponding to the final EOS token to produce the embedding.
Qwen3 provides both embedding and reranking models, including 0.6B, 4B and 8B variants.
A leaderboard position is not the same thing as universal superiority because embedding performance varies across tasks.

Qwen3 Embedding: Advancing Text Embedding and Reranking with Foundation Models

arxiv.org2025-06-11authorityabout the Qwen3 Embedding familyHN 1 pts · 0 comments433 words

The Qwen3 Embedding series introduces new text embedding and reranking models built on Qwen3 foundation models. It offers a spectrum of model sizes (0.6B, 4B, 8B) for both embedding and reranking tasks and is publicly available under the Apache 2.0 license. The series achieves strong results on the multilingual evaluation benchmark MTEB for text embedding.written, reviewed, cited

Fine-Tuning Qwen3 Embeddings for product category classification

blog.ivan.digital2026-03-03about the Qwen3 Embedding familyHN 2 pts · 0 comments

Qwen3-Embedding-8B Retriever

emergentmind.com1,144 words

Qwen3-Embedding-8B Retriever is an 8-billion parameter dense text embedding model with 36 decoder blocks that converts queries and documents into 4096 floating-point vectors. The model uses a two-stage contrastive training pipeline and achieves per-encode latency in the 300–400 ms range on A100/H100 GPUs for deployment.written, reviewed, cited

The model merges several fine-tuning checkpoints using spherical linear interpolation to improve robustness and cross-domain generalization.
Reasoning-intensive performance increases from an nDCG@10 of 22.8 to 38.1 after applying LoRA-based adapters with Redapter self-adaptive loss.
The 4B variant achieves over 95% of the 8B model's accuracy while consuming only about 60% of its memory and latency footprint.

Qwen3-Embedding-8B: Multilingual Text Embeddings

emergentmind.com1,286 words

Qwen3-Embedding-8B/4096 Model Overview

emergentmind.com1,202 words

QwenLM/Qwen3-Embedding

github.com2025-06-05authorityabout the Qwen3 Embedding family2,980 words

Alibaba Qwen Team Releases Qwen3-Embedding and Qwen3-Reranker Series – Redefining Multilingual Embedding and Ranking Standards - MarkTechPost

marktechpost.com2025-06-06about the Qwen3 Embedding family

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models this model

Yanzhao Zhang, Mingxin Li, Dingkun Long +9 · 2025-06-05 · PDF · HF paper page (86 upvotes)

Similar models

3 models

Similar models

bge-multilingual-gemma2 · 9.24B
same task (feature-extraction), fewer providers
Qwen3-Reranker-8B · 8.19B
same size band (M), fewer providers, permissive license
bge-large-zh-v1.5
same task (feature-extraction), fewer providers, permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:33 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-25

Follow Qwen3-Embedding-8B

Watch this model

/feeds/models/Qwen--Qwen3-Embedding-8B.xml(RSS 2.0) · every change to Qwen3-Embedding-8B
/feeds/families/Qwen--Qwen3-Embedding.xml(RSS 2.0) · every change across the Qwen3-Embedding family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.