Up to date
Models › Qwen3-Embedding family
Qwen Apache 2.0 out 10 months ago
8.14B parameters, text in, embeddings out, 256K tokens of context.
Verified 2026-10-25 Rank #35 in Embeddings · score 92.9 method v1 Sources 3 Downloads 30d 1.3M · #96 of 619 tracked
Changed this week: 1 new quantized build on the Hub (22 in all).2026-10-24
Picked by rule from 3 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The Qwen3 Embedding series introduces new text embedding and reranking models built on Qwen3 foundation models.
The 1 pick above, plus these 2. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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.
No API provider serves it, so its downloads are people running it themselves: 22 quantized builds and 7 Spaces.
Top Spaces running it: baohuynhbk14/Qwen3-VL-Multimodal-Search-DEMO ♥ 17 · nanovdr/distilling-the-document-tower ♥ 2
Weights only, an estimate: leave room for the KV cache. Method
About 4.9 GB at INT4, 9.8 GB at FP8, 20 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-01-07), amber triangles are 3 dated articles (each links to it), grey ticks are repo commits.
29 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported MMEB · Embed text for search | 74.4 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 81 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 80 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 92.2 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 80.1 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 79.1 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 70.1 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 57 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 53.2 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 66.1 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 88.2 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 69.9 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 88.8 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 78.3 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 83.3 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MMEB · Embed text for search | 77.9 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 67.88 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 58.88 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 77.48 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 71.95 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 55.82 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 4.46 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 28.59 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 81.08 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 65.72 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported MTEB · Embed text for search | 69.41 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 75.41 | Qwen3-VL-Embedding-8B | model author | 2026-10-07 |
The model is designed for multimodal information retrieval and supports text, images, screenshots, videos, and mixed modality inputs.[1] It is released under the Apache 2.0 license and supports over 30 languages.[2][3]
Generated 2026-10-07 15:02 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
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.
Listed in Embeddings
| Pipeline tag | sentence-similarity |
|---|---|
| Library | sentence-transformers |
| Architectures | Qwen3VLForConditionalGeneration |
| Model type | qwen3_vl |
| Dtype | bfloat16 |
| Params by dtype | BF16: 8,144,793,840 |
| 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 | 2026-10-25 |
vLLM: supported · Qwen3VLForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3VLForConditionalGeneration · 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.
Weights-only memory for Qwen3-VL-Embedding-8B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 20 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 9.8 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 4.9 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Base model: Qwen/Qwen3-VL-8B-Instruct
quantized: 22finetune: 6adapter: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| RamManavalan/Qwen3-VL-Embedding-8B-FP8 | quantized | -FP8 | 8.77B | M | apache-2.0 | RamManavalan | 6 | 1.2M |
| VesNFF/Qwen3-VL-Embedding-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | VesNFF | 0 | 206.2k |
| mradermacher/Qwen3-VL-Embedding-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 10 | 32.6k |
| collin-park/Qwen3-VL-Embedding-8B-W8A8 | quantized | -W8A8 | 8.77B | M | apache-2.0 | collin-park | 1 | 30.6k |
| sywang/TPIPS-ActDiff-Qwen3VL-8B | finetune | — | — | — | other | sywang | 1 | 11.7k |
| dam2452/Qwen3-VL-Embedding-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | dam2452 | 4 | 11.2k |
| sywang/TPIPS-Embed-Qwen3VL-8B | finetune | — | — | — | other | sywang | 2 | 11.1k |
| sywang/TPIPS-EarlyFusion-Qwen3VL-8B | finetune | — | — | — | other | sywang | 1 | 10.7k |
| gonuit/Qwen3-VL-Embedding-8B-AWQ-4bit | quantized | AWQ | 2.13B | M | apache-2.0 | gonuit | 0 | 10.6k |
| batiai/Qwen3-VL-Embedding-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | batiai | 7 | 7.9k |
| mradermacher/Qwen3-VL-Embedding-8B-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 1 | 7.5k |
| aiteza/Qwen3-VL-Embedding-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | aiteza | 5 | 4.2k |
| andrzejmontano/Qwen3-VL-Embedding-8B-mlx-8bit | quantized | mlx | 8.77B | M | apache-2.0 | andrzejmontano | 0 | 3.5k |
| jedisct1/Qwen3-VL-Embedding-8B-mlx | quantized | mlx | 8.19B | M | apache-2.0 | jedisct1 | 0 | 3.3k |
| Ganonook/Qwen3-VL-Embedding-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | Ganonook | 3 | 2.6k |
| nkamiy/Qwen3-VL-Embedding-8B-8bit-mlx | quantized | -8bit | 8.77B | M | apache-2.0 | nkamiy | 0 | 2.6k |
| QuaduxIT/Qwen3-VL-Embedding-8B-W8A16 | quantized | — | 3.56B | M | apache-2.0 | QuaduxIT | 0 | 2.4k |
| Zeknes/Qwen3-VL-Embedding-8B-MLX-4bit | quantized | MLX | 8.77B | M | apache-2.0 | Zeknes | 0 | 1.8k |
| shinshekai/Qwen3-VL-Embedding-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | shinshekai | 0 | 1.2k |
| lainsoykaf/Qwen3-VL-Embedding-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | lainsoykaf | 0 | 905 |
| OLAIR/OLA-Embed | finetune | — | 8.14B | M | apache-2.0 | OLAIR | 0 | 507 |
| ganeshrao/Qwen3-VL-Embedding-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | ganeshrao | 0 | 481 |
| gorse-io/Qwen3-VL-Embedding-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | gorse-io | 0 | 415 |
| ganeshrao/Qwen3-VL-Embedding-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | ganeshrao | 0 | 408 |
| indifferentketchup/Qwen3-VL-Embedding-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | indifferentketchup | 0 | 332 |
| hbaighdc/Qwen3-VL-Embedding-8B-Q3_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | hbaighdc | 0 | 231 |
| lloydchristmasx/Qwen3-VL-Embedding-8B | finetune | — | — | — | apache-2.0 | lloydchristmasx | 0 | 195 |
| woooooz/Qwen3-VL-Embedding-8B | finetune | — | 8.14B | M | apache-2.0 | woooooz | 0 | 21 |
| ericssonbear/qwen3-vl-emb-8b-mrright-zhtw-lora | adapter | — | — | — | apache-2.0 | ericssonbear | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2026-04-16 | Integrate with Sentence Transformers v5.4 (#11) | 2c45655 |
| 2026-01-19 | add model pipeline info and technical report info (#8) | 418d038 |
| 2026-01-09 | Add vLLM and SGLang usage examples (#4) | a12d611 |
| 2026-01-08 | Update scripts/qwen3_vl_embedding.py (#3) | 93dc85c |
| 2026-01-08 | Update README.md (#2) | 54edbcd |
| 2026-01-08 | Update README.md | 5407fcd |
| 2026-01-07 | Update README.md | db0a90a |
| 2026-01-07 | Upload folder using huggingface_hub | 402b6f3 |
| 2026-01-07 | initial commit | 0115b0f |
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.
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).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:35 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:02 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
