multi-qa-MiniLM-L6-cos-v1

sentence-transformers Not declared out 5 years ago

22.7M parameters, text in, embeddings out, 512 tokens of context. 1 provider serves it.

Verified 2026-10-26 07:27 UTC Rank #56 in Embeddings · score 87.8 method v1 Sources 3 Downloads 30d 603.9k · #147 of 619 tracked

Can I use it?Not declaredNo license declared on the model cardlicence file2026-10-26

Does it work?67.83on MTEB#9 of 64 measuredleaderboard2026-10-24

Do people use it?603.9kdownloads, 30 days#147 of 619 we track · top 24%+0 likes, downloads −10% this week138 likesHugging Face2026-10-26

Worth reading

Picked by rule from 5 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 687 HN pts

Accelerating Gemma 4: faster inference with multi-token prediction drafters

Google released Multi-Token Prediction (MTP) drafters for the Gemma 4 family under the same open-source Apache 2.0 license as Gemma 4.

The MTP drafter shares the target model's KV cache to avoid recalculating context the larger model has already processed.
For E2B and E4B edge models, an efficient clustering technique in the embedder accelerates generation where logit calculation is a bottleneck.

blog.google 2026-05-05 687 HN pts · 330 comments 796 words about the multi family family

Field report

Using multi-qa-minilm-l6-cos-v1

A Kaggle notebook demonstrates using the multi-qa-minilm-l6-cos-v1 model for open book Q&A by appending Wikipedia context to test samples.

The notebook installs transformers version 4.31.0 by uninstalling the existing 4.30.2 version.

kaggle.com 3k words

5 best of 5 tracked

The 3 picks above, plus these 2. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

Articlemulti-qa-MiniLM-L6-cos-v1 by Sentence Transformerssavrn.com · undated
ArticleMulti Qa MiniLM L6 Cos V1 · Models · Dataloopdataloop.ai · undated

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; 4 quantized builds for local use.

As a service1HF Inference
On own hardware4quantized builds
Built on3530 fine-tunes · 1 adapters · 4 merges
In apps88Hugging Face Spaces

Top Spaces running it: jorge-henao/ask2democracycol ♥ 5 · Abhilashvj/haystack_QA ♥ 5 · namoopsoo/text-audio-video ♥ 4 · Sukhyun/course_recommender ♥ 2

Will it fit?

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

Fits a 16 GB card at BF16

About 0.1 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.

#9 of 64 on MTEB so far

67.83. Each tick is a model on this site with a measured MTEB row; author-reported numbers stay in Reference. Mean over 19/41 tasks present. 19/41 tasks

Source · leaderboard · 2026-10-24

The story so far

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

1 dated article and 25 repo commits so far

Teal is the weights (2022-03-02), amber triangles are 1 dated article (each links to it), grey ticks are repo commits.

What people built on it

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

39 derivatives on the Hub, most of them fine-tunes

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

GGUF Felladrin/gguf-multi-qa-MiniLM-L6-cos-v1 for llama.cpp, Ollama, LM Studio 5.1k downloads

All variants →

Reference

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

Benchmarks

2 rows

Benchmarks 2 rows

measured third-party leaderboard row (2)author-reported number from the model card or eval filesNo 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	67.83 mean over 19/41 tasks present	19/41 tasks	leaderboard	2026-10-24
measured MTEB Reranking · Rerank search results	46.99 MAP x100 (mean of 2 tasks)	2 tasks, split=test	leaderboard	2026-10-24

What the card says

2 cited sentences

Summary generated, cited

The multi-qa-MiniLM-L6-cos-v1 is a BERT-based model with 22.7M parameters designed for semantic search.^[1]^[2]^[3] It maps sentences and paragraphs to a 384 dimensional dense vector space.^[4]

parameters human
architectures
“designed for **semantic search**”
“maps sentences & paragraphs to a 384 dimensional dense ve…”

Generated 2026-10-07 15:04 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

2 use cases

What it is good at 2 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

20 fields

Listed in Embeddings

Facts

multi-qa-MiniLM-L6-cos-v1 has 22.7M parameters. (safetensors metadata, observed 2026-10-26)
Context length is 512 tokens (512). (config.json max_position_embeddings, observed 2026-10-26)
License: Not declared (friction: review). No license declared on the model card. (license file, observed 2026-10-26)
Created 2022-03-02, last modified 2024-11-05. (HF API, observed 2026-10-26)

Pipeline tag	sentence-similarity
Library	sentence-transformers
Architectures	BertModel
Model type	bert
Params by dtype	F32: 22,713,216 · I64: 512
Size band	S
MoE	no
Context length	512 (config.json max_position_embeddings)
Languages	en
License	Not declared
Hub snapshot	2026-10-26

Servability

vLLM · SGLang

Servability

vLLM: supported · BertModel · 4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60SGLang: supported · BertModel · 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 multi-qa-MiniLM-L6-cos-v1: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	0.1 GB · Edge (<4 GB)
FP8 (1 byte/param)	0.0 GB · Edge (<4 GB)
INT4 (0.5 bytes/param)	0.0 GB · Edge (<4 GB)

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

All variants

39 variants

Lineage

Original model Family: sentence-transformers/multi 39 derivatives on the Hub

finetune: 30quantized: 4merge: 4adapter: 1

Top variants (39 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
ntproctor/mass-academy-faq-embedder	finetune	—	22.7M	S	—	ntproctor	0	25.8k
Xenova/multi-qa-MiniLM-L6-cos-v1	quantized	—	—	—	—	Xenova	3	13.3k
Felladrin/gguf-multi-qa-MiniLM-L6-cos-v1	quantized	gguf	—	—	—	Felladrin	0	5.1k
mradermacher/multi-qa-MiniLM-L6-cos-v1-i1-GGUF	quantized	GGUF	—	—	—	mradermacher	0	4.6k
mradermacher/multi-qa-MiniLM-L6-cos-v1-GGUF	quantized	GGUF	—	—	—	mradermacher	0	2.3k
clairedhx/autotrain-transcriptions	finetune	—	22.7M	S	—	clairedhx	0	1.1k
clairedhx/autotrain-phe-avis-ct	finetune	—	22.7M	S	—	clairedhx	0	859
Baiming123/Calcu_Disease_Similarity	finetune	—	—	—	—	Baiming123	3	553
Trelis/multi-qa-MiniLM-L6-cos-v1-ft-pairs-1-epoch-scale-20	finetune	—	22.7M	S	—	Trelis	0	519
MohamedAshraf701/multi-qa-MiniLM-L6-cos-v1-products	finetune	—	22.7M	S	mit	MohamedAshraf701	2	339
Trelis/multi-qa-MiniLM-L6-cos-v1-ft-pairs	finetune	—	22.7M	S	—	Trelis	0	436
Trelis/multi-qa-MiniLM-L6-cos-v1-ft-pairs-1-epoch	finetune	—	22.7M	S	—	Trelis	0	427
bau0221/ptz_embedding	finetune	—	22.7M	S	—	bau0221	0	418
guipa01/telos-demo-model	finetune	—	22.7M	S	—	guipa01	0	405
SwapnilVishwakarma/USMED	finetune	—	22.7M	S	—	SwapnilVishwakarma	0	379
guipa01/best-telos-model	finetune	—	22.7M	S	—	guipa01	0	377
bau0221/ptz_embedding_2	finetune	—	22.7M	S	—	bau0221	0	349
hutuhehe/finretriever-chunk-retriever	finetune	—	22.7M	S	—	hutuhehe	0	324
hutuhehe/finretriever-doc-type-ranker	finetune	—	22.7M	S	—	hutuhehe	0	319
fabioaloisio/multi-qa-MiniLM-L6-cos-v1-hydro-ft-pairs	finetune	—	22.7M	S	—	fabioaloisio	0	314
recruitco/embedding_criteria_profile_summary_matching_qa_minilm_v1	finetune	—	22.7M	S	—	recruitco	0	304
Boltuzamaki/multi-qa-MiniLM-L6-cos-v1-Tough_Rugby_ft	finetune	—	22.7M	S	—	Boltuzamaki	0	296
Trelis/multi-qa-MiniLM-L6-cos-v1-ft-pairs-2-cos-epoch-s20	finetune	—	22.7M	S	—	Trelis	0	264
poojavalecha09/sentence-transformer-finetuned-faq	finetune	—	22.7M	S	—	poojavalecha09	0	203
Aathi13/finetuned-biencoder-campusbot	finetune	—	22.7M	S	—	Aathi13	0	183
anomys/gsm-finetunned-v3	finetune	—	22.7M	S	—	anomys	0	134
howanching-clara/all_keywords_multi-qa-MiniLM-L6-cos-v1_final	finetune	—	22.7M	S	—	howanching-clara	0	117
LamaDiab/MultiMiniLM-V25Data-256BATCH-SemanticEngine	finetune	—	22.7M	S	—	LamaDiab	0	111
howanching-clara/all_abstracts_multi-qa-MiniLM-L6-cos-v1_final	finetune	—	22.7M	S	—	howanching-clara	0	108
Bombek1/multi-qa-MiniLM-L6-cos-v1-litert	finetune	—	—	—	apache-2.0	Bombek1	0	87
yjoonjang/all-MiniLM-L6-v2-multi-qa-slerp-merge	merge	—	22.7M	S	—	yjoonjang	0	73
yjoonjang/all-MiniLM-L6-v2-multi-qa-linear-merge	merge	—	22.7M	S	—	yjoonjang	0	70
aneforge/multi-qa-MiniLM-L6-cos-v1	finetune	—	22.7M	S	—	aneforge	0	63
yjoonjang/all-MiniLM-L6-v2-multi-qa-dare-ties-merge	merge	—	22.7M	S	—	yjoonjang	0	62
yjoonjang/all-MiniLM-L6-v2-multi-qa-ties-merge	merge	—	22.7M	S	—	yjoonjang	0	57
rellanchayan/QLoRAEmbeddingFineTune	adapter	—	—	—	—	rellanchayan	0	33
zeromodels/multi-qa-MiniLM-L6-cos-v1	finetune	—	—	—	—	zeromodels	0	32
Fred-William/african-medical-qa-assistant	finetune	—	22.7M	S	cc-by-nc-sa-4.0	Fred-William	0	27
Jagadesheeden/RAG_PDF	finetune	—	—	—	—	Jagadesheeden	0	0

Repository changelog

25 commits

Repository changelog 25 commits

Date	Commit	SHA
2024-11-05	Add exported openvino model 'openvino_model_qint8_quantized.xml'	b207367
2024-10-11	Add new SentenceTransformer model with an openvino backend	2d981ed
2024-10-11	Add new SentenceTransformer model with an openvino backend	e77d999
2024-10-11	Add exported ONNX model 'model_qint8_avx512_vnni.onnx'	8ed8b01
2024-10-11	Add exported ONNX model 'model_qint8_avx512.onnx'	72251d4
2024-10-11	Add exported ONNX model 'model_quint8_avx2.onnx'	9c441fc
2024-10-11	Add exported ONNX model 'model_qint8_arm64.onnx'	3454cde
2024-10-11	Add exported ONNX model 'model_O4.onnx'	fa5d39a
2024-10-11	Add exported ONNX model 'model_O3.onnx'	cd42685
2024-10-11	Add exported ONNX model 'model_O2.onnx'	49b099c
2024-10-11	Add exported ONNX model 'model_O1.onnx'	c9cafa9
2024-10-11	Add new SentenceTransformer model with an onnx backend	40426cd
2024-03-27	Adding `safetensors` variant of this model (#5)	2430568
2024-02-12	Update Sentence Transformers metadata (#7)	3884516
2023-11-02	Update README.md (#6)	c7a0b07
2022-07-11	adding training dataset	2ad254d
2022-07-08	add training datasets	0de4dba
2022-07-08	Update README.md	ccac161
2022-07-07	adding training datasets	b52b070
2022-06-21	Model Card with TensorFlow example (#1)	491739e
2022-06-15	Add TF weights	e8717c2
2021-08-23	update	1de2325
2021-08-23	update	7ba36cb
2021-08-23	upload	ad23f13
2021-08-23	initial commit	c210786

All providers

1 provider · 1 endpoint

Where to run 1 providers · 1 endpoints

Hosted endpoints observed for multi-qa-MiniLM-L6-cos-v1, 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	sentence-transformers/multi-qa-MiniLM-L6-cos-v1	—	—	—	—	—	—	—	live	2026-10-26

Articles and coverage

5 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

Accelerating Gemma 4: faster inference with multi-token prediction drafters

blog.google2026-05-05authorityHN front pageabout the multi familyHN 687 pts · 330 comments796 words

Google released Multi-Token Prediction (MTP) drafters for the Gemma 4 family under the same open-source Apache 2.0 license as Gemma 4. These drafters deliver up to a 3x speedup for models like the 26B MoE and 31B Dense versions without quality degradation. Benchmarks on hardware such as the NVIDIA RTX PRO 6000 show that increasing batch sizes to 4 to 8 unlocks up to a ~2.2x speedup locally.written, reviewed, cited

The MTP drafter shares the target model's KV cache to avoid recalculating context the larger model has already processed.
For E2B and E4B edge models, an efficient clustering technique in the embedder accelerates generation where logit calculation is a bottleneck.
The 26B mixture-of-experts model presents unique routing challenges at a batch size of 1 on Apple Silicon.

Analysis4

multi-qa-MiniLM-L6-cos-v1 by Sentence Transformers

savrn.com1,683 words

multi-qa-MiniLM-L6-cos-v1 is an open-weight sentence similarity model from Sentence Transformers designed for semantic search. It has 23M parameters, a 512-token context, and requires about 0.1 GB of GPU memory at 16-bit precision. The model was released March 2, 2022, and fits on a 1x MI300X from Vultr at $1.85 an hour.written, reviewed, cited

The model license field is blank, meaning written terms are required before any revenue-generating deployment ships.
Memory usage rounds to 0.0 GB at 8-bit and 4-bit precisions, though a long context requires additional space.
The model was trained on 215M question-and-answer pairs from diverse sources including ms_marco, gooaq, and eli5.

all-MiniLM-L6-v2 vs multi-qa-MiniLM-L6-cos-v1: Specs and GPU Cost

savrn.com842 words

all-MiniLM-L6-v2 and multi-qa-MiniLM-L6-cos-v1 are 23M parameter sentence transformers requiring 0.1 GB memory at 16-bit. all-MiniLM-L6-v2 carries an Apache 2.0 license, whereas multi-qa-MiniLM-L6-cos-v1 has no stated license despite open access. The latter was trained on 215M question-and-answer pairs and both models run on a single MI300X GPU for $1.85 per hour.written, reviewed, cited

The multi-qa-MiniLM-L6-cos-v1 license field is blank, meaning open access does not grant permission for commercial deployment without written terms.
The all-MiniLM-L6-v2 model lineage traces back to nreimers/MiniLM-L6-H384-uncased as the source model.
The multi-qa-MiniLM-L6-cos-v1 was released on March 2, 2022, and last updated on November 5, 2024.

Using multi-qa-minilm-l6-cos-v1

kaggle.com3,005 words

A Kaggle notebook demonstrates using the multi-qa-minilm-l6-cos-v1 model for open book Q&A by appending Wikipedia context to test samples. The approach combines a trained Q&A model with Mgoksu's original model in a 50% 50% ensemble to infer multiple choice answers, achieving a private score of 0.788006 and a best score of 0.799208. Context is clipped to 1500 characters to prevent out-of-memory errors during inference.written, reviewed, cited

The notebook installs transformers version 4.31.0 by uninstalling the existing 4.30.2 version.

Multi Qa MiniLM L6 Cos V1 · Models · Dataloop

dataloop.ai

Similar models

3 models

Similar models

paraphrase-MiniLM-L6-v2 · 22.7M
same size band (S), same task (sentence-similarity), more providers, permissive license
e5-small-v2 · 33.4M
same size band (S), same task (sentence-similarity), permissive license
bge-small-zh-v1.5 · 24.0M
same size band (S), 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 15:04 UTC · Last verified 2026-10-26 07:27 UTC

Follow multi-qa-MiniLM-L6-cos-v1

Watch this model

/feeds/models/sentence-transformers--multi-qa-MiniLM-L6-cos-v1.xml(RSS 2.0) · every change to multi-qa-MiniLM-L6-cos-v1
/feeds/families/sentence-transformers--multi.xml(RSS 2.0) · every change across the multi family

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