Qwen3-VL-Reranker-2B

2.13B parameters, a query and passages in, relevance scores out, 256K tokens of context. 1 provider serves it.

Verified 2026-10-26 07:27 UTC Rank #54 in Embeddings · score 88.5 method v1 Sources 3 Downloads 30d 94.8k · #318 of 619 tracked

Can I run it?1.3GB at INT4Estimatefits a 16 GB cardmethod

Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-26

Do people use it?94.8kdownloads, 30 days#318 of 619 we track · below the median237 likesHugging Face2026-10-26

Worth reading

Picked by rule from 1 tracked piece: the lab's own post, the most shared, a field report, the latest.

Qwen3VL-Reranker-2B achieves a maximum input resolution of 1280 tokens per images, almost twice the resolution of the author's model.

From Qwen3-VL-Reranker-2B VS Qwen3VLRerankerMM-2B : Lightweight vs Industrial Multimodal Reranking with Qwen 3 · huggingface.co · 2026-01-12

Latest · 9 months ago

Qwen3-VL-Reranker-2B VS Qwen3VLRerankerMM-2B : Lightweight vs Industrial Multimodal Reranking with Qwen 3

Qwen released the Qwen3VLRerankerMM-2B in January 2026, trained on 40 million pairs using all model weights including the LM head.

The author's model uses LoRA while keeping the LM head frozen to prevent overfitting and preserve robustness.
Qwen3VL-Reranker-2B achieves a maximum input resolution of 1280 tokens per images, almost twice the resolution of the author's model.

huggingface.co 2026-01-12 1.1k words

Where to run it

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

#	Cheapest live row	In /1M	Out /1M	Context	Quant	Seen
1	HF Inference	—	—	—	—	Oct 26

All 1 provider →

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 1 API provider; 18 quantized builds for local use.

As a service1HF Inference
On own hardware18quantized builds
Built on54 fine-tunes · 1 adapters · 0 merges
In apps10Hugging Face Spaces

Top Spaces running it: baohuynhbk14/Qwen3-VL-Multimodal-Search-DEMO ♥ 17 · localellm/WebGPU-qwen3-vl-reranker-2b-i1-gguf ♥ 2

Will it fit?

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

Fits a 16 GB card at INT4

About 1.3 GB at INT4, 2.6 GB at FP8, 5.1 GB at BF16: weights only, so leave room for the KV cache.

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

The story so far

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

1 dated article and 9 repo commits so far

Teal is the weights (2026-01-07), amber triangles are 1 dated article (each links to it), grey ticks are repo commits.

What people built on it

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

23 derivatives on the Hub, most of them quantized

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

GGUF mradermacher/Qwen3-VL-Reranker-2B-i1-GGUF for llama.cpp, Ollama, LM Studio 173k downloads MLX starkdmi/Qwen3-VL-Reranker-2B-mlx-5bit for Apple silicon 359 downloads NVFP4 jeffpeng3/Qwen3-VL-Reranker-2B-NVFP4 for vLLM, SGLang 5.2k downloads

All variants →

Reference

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

Benchmarks

7 rows

Benchmarks 7 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (7)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
author-reported JinaVDR	80.9	Qwen3-VL-Reranker-2B	model author	2026-10-07
author-reported MMEB · Embed text for search	75.1	Qwen3-VL-Reranker-2B	model author	2026-10-07
author-reported MMEB · Embed text for search	73.8	Qwen3-VL-Reranker-2B	model author	2026-10-07
author-reported MMEB · Embed text for search	52.1	Qwen3-VL-Reranker-2B	model author	2026-10-07
author-reported MMEB · Embed text for search	83.4	Qwen3-VL-Reranker-2B	model author	2026-10-07
author-reported MTEB · Embed text for search	70	Qwen3-VL-Reranker-2B	model author	2026-10-07
author-reported ViDoRe · Embed text for search	60.8	Qwen3-VL-Reranker-2B	model author	2026-10-07

What the card says

3 cited sentences

Summary generated, cited

Qwen3-VL-Reranker-2B is a 2.13B parameter multimodal rerank model fine-tuned from the Qwen/Qwen3-VL-2B-Instruct base model.^[1]^[2] The model processes text, images, screenshots, and videos to output relevance scores for (query, document) pairs.^[3] It is released under the Apache 2.0 license and supports over 30 languages.^[4]^[5]

parameters human
“base_model:finetune:Qwen/Qwen3-VL-2B-Instruct”
“Supported Input Modalities: Text, images, screenshots, vi…”
license
“Supported Languages: 30+ Languages”

Generated 2026-10-07 15:05 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

9 use cases

What it is good at 9 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.

Facts

21 fields

Listed in Embeddings

Facts

Qwen3-VL-Reranker-2B has 2.13B parameters. (safetensors metadata, observed 2026-10-26)
Context length is 262,144 tokens (256K). (config.json max_position_embeddings, observed 2026-10-26)
License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-26)
Created 2026-01-07, last modified 2026-04-16. (HF API, observed 2026-10-26)

Pipeline tag	text-ranking
Library	transformers
Architectures	Qwen3VLForConditionalGeneration
Model type	qwen3_vl
Dtype	bfloat16
Params by dtype	BF16: 2,127,532,032
Size band	M
MoE	no
Context length	262,144 (config.json max_position_embeddings)
Papers	arXiv:2601.04720
License	Apache 2.0 (apache-2.0)
Hub snapshot	2026-10-26

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen3VLForConditionalGeneration · 4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60SGLang: supported · Qwen3VLForConditionalGeneration · 60d941293e3af27a12c6eb64e2c3e2e1a17f100f

vLLM registry (4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60) · SGLang registry (60d941293e3af27a12c6eb64e2c3e2e1a17f100f) · 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-VL-Reranker-2B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	5.1 GB · Single GPU (<24 GB)
FP8 (1 byte/param)	2.6 GB · Edge (<4 GB)
INT4 (0.5 bytes/param)	1.3 GB · Edge (<4 GB)

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

All variants

23 variants

Lineage

Fine-tuned from Qwen/Qwen3-VL-2B-Instruct (same organization) Family: Qwen/Qwen3-Reranker 23 derivatives on the Hub

Base model: Qwen/Qwen3-VL-2B-Instruct

quantized: 18finetune: 4adapter: 1

Top variants (23 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
mradermacher/Qwen3-VL-Reranker-2B-i1-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	2	173k
mradermacher/Qwen3-VL-Reranker-2B-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	8	55k
jeffpeng3/Qwen3-VL-Reranker-2B-NVFP4	quantized	NVFP4	1.51B	S	apache-2.0	jeffpeng3	1	5.2k
bealore/Qwen3-VL-Reranker-2B-GGUF	quantized	GGUF	—	—	apache-2.0	bealore	0	2.4k
nlpai-lab/RenderRank-2B	finetune	—	2.13B	M	apache-2.0	nlpai-lab	16	331
tooktang/Qwen3-VL-Reranker-2B-Q8_0-GGUF	quantized	Q8_0	—	—	apache-2.0	tooktang	0	696
MrMati/Qwen3-VL-Reranker-2B-GGUF	quantized	GGUF	—	—	apache-2.0	MrMati	0	641
Surpem/Supertron2-Reranker-2B	finetune	—	2.13B	M	apache-2.0	Surpem	5	119
AhmedxSaad/Qwen3-VL-Reranker-2B-Q4_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	AhmedxSaad	1	315
starkdmi/Qwen3-VL-Reranker-2B-mlx-5bit	quantized	mlx	2.13B	M	apache-2.0	starkdmi	0	359
staralt/Qwen3-VL-Reranker-2B-Q4_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	staralt	0	322
imbenwolf/Qwen3-VL-Reranker-2B-Q4_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	imbenwolf	0	296
turingevo/Qwen3-VL-Reranker-2B-Q4_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	turingevo	0	275
staralt/Qwen3-VL-Reranker-2B-Q5_K_M-GGUF	quantized	GGUF	—	—	apache-2.0	staralt	0	267
staralt/Qwen3-VL-Reranker-2B-IQ4_NL-GGUF	quantized	GGUF	—	—	apache-2.0	staralt	0	169
staralt/Qwen3-VL-Reranker-2B-Q8_0-GGUF	quantized	Q8_0	—	—	apache-2.0	staralt	0	169
gorse-io/Qwen3-VL-Reranker-2B-GGUF	quantized	GGUF	—	—	apache-2.0	gorse-io	0	164
soyrsoyr/Qwen3-VL-Reranker-0.1B-tiny	finetune	—	98.7M	S	—	soyrsoyr	0	109
t-firefly/qwen3-vl-reranker-2b-rknn3-rk1820	quantized	GGUF	—	—	apache-2.0	t-firefly	0	64
t-firefly/qwen3-vl-reranker-2b-rknn3-rk1828	quantized	GGUF	—	—	apache-2.0	t-firefly	0	49
inference-optimization/Qwen3-VL-Reranker-0.6B-small	finetune	—	592M	S	apache-2.0	inference-optimization	0	38
hungnghehe/qwen3-vl-reranker-2b-ucf-rerank-phase1-v6	adapter	—	—	—	—	hungnghehe	0	26
r3lax/Qwen3-VL-Reranker-2B-ZOIDMATTER-NVFP4	quantized	NVFP4	1.42B	S	apache-2.0	r3lax	0	18

Repository changelog

9 commits

changed in the last 7 daysNew on the watch list: 2026-10-26all changes

Repository changelog 9 commits

Date	Commit	SHA
2026-04-16	Integrate with Sentence Transformers v5.4 (#11)	4bd860a
2026-01-19	add model pipeline info and technical report info (#10)	9887917
2026-01-09	Add basic vllm usage (#6)	76219da
2026-01-08	Update README.md (#3)	00002d1
2026-01-08	Update README.md (#2)	75a922c
2026-01-08	Update README.md	e3511d3
2026-01-07	Update README.md	1b97955
2026-01-07	Upload folder using huggingface_hub	f9ca7f1
2026-01-07	initial commit	b64a8fd

All providers

1 provider · 1 endpoint

Where to run 1 providers · 1 endpoints

Hosted endpoints observed for Qwen3-VL-Reranker-2B, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
HF Inference hf	Qwen/Qwen3-VL-Reranker-2B	—	—	—	—	—	—	—	live	2026-10-26

Articles and coverage

1 piece

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-VL-Reranker-2B VS Qwen3VLRerankerMM-2B : Lightweight vs Industrial Multimodal Reranking with Qwen 3

huggingface.co2026-01-12authority1,095 words

Qwen released the Qwen3VLRerankerMM-2B in January 2026, trained on 40 million pairs using all model weights including the LM head. The model handles text, image, or even a video for both queries and documents, unlike cross-modal approaches limited to text queries and image documents. Executing the language model head during the forward pass represents roughly 14% of the FLOPs in a single forward pass.written, reviewed, cited

The author's model uses LoRA while keeping the LM head frozen to prevent overfitting and preserve robustness.
Qwen3VL-Reranker-2B achieves a maximum input resolution of 1280 tokens per images, almost twice the resolution of the author's model.
The reranker was trained on 40 million pairs, both synthetic and collected, using synthetic data for videos.

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-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking this model

Mingxin Li, Yanzhao Zhang, Dingkun Long +9 · 2026-01-08 · PDF · HF paper page (59 upvotes)

Similar models

3 models

Similar models

Qwen3-VL-Embedding-2B · 2.13B
same size band (M), permissive license
CLM-v0.1-8B
same task (text-ranking), permissive license
Qwen3-Embedding-4B · 4.02B
same size band (M), permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-26 · HF API: 2026-10-26 07:18 UTC · Providers: 2026-10-26 07:26 UTC · Articles: 2026-10-26 07:27 UTC · Summary: 2026-10-07 15:05 UTC · Last verified 2026-10-26 07:27 UTC

Follow Qwen3-VL-Reranker-2B

Watch this model

/feeds/models/Qwen--Qwen3-VL-Reranker-2B.xml(RSS 2.0) · every change to Qwen3-VL-Reranker-2B
/feeds/families/Qwen--Qwen3-Reranker.xml(RSS 2.0) · every change across the Qwen3-Reranker family

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