{
    "meta": {
        "title": "D-Central — Local Embedding Models for RAG",
        "description": "Self-hostable text-embedding models (20) by embedding dimensions, max context, MTEB score (benchmark named), VRAM, Ollama tag, license and multilingual coverage — the retrieval layer of a self-hosted RAG stack.",
        "generated": "2026-07-17T17:56:57+00:00",
        "last_verified": "July 2026",
        "version": "1.0",
        "license": "https://creativecommons.org/licenses/by/4.0/",
        "license_name": "CC BY 4.0",
        "source": "https://d-central.tech/local-embedding-models/",
        "record_count": 20,
        "provenance": "MTEB leaderboard (HuggingFace mteb/leaderboard) + each model's HuggingFace card + the Ollama library, reconciled adversarially. Ollama tags verified against the live library; MTEB scores labelled with their exact benchmark (English MTEB v2 and MMTEB are not directly comparable).",
        "disclaimer": "MTEB scores are only comparable WITHIN the same benchmark; check the benchmark column. VRAM is approximate for typical local precision and excludes long-context KV overhead. Confirm the current model card + license before production use."
    },
    "rows": [
        {
            "model_id": "sentence-transformers/all-MiniLM-L6-v2",
            "model_name": "all-MiniLM-L6-v2",
            "family": "MiniLM L6 (sentence-transformers)",
            "developer": "sentence-transformers",
            "params": "22.7M",
            "embedding_dimensions": 384,
            "matryoshka_dims": null,
            "max_context_tokens": 256,
            "mteb_score": 56.25999999999999801048033987171947956085205078125,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean ~56.26 (leaderboard, NOT on card; card lists only ArguAna 50.17)",
            "approx_vram_gb": 0.1000000000000000055511151231257827021181583404541015625,
            "ollama_tag": "all-minilm",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": false,
            "lang_count": "1 (English)",
            "best_for": "Ultra-light CPU-only RAG, prototyping, very high throughput; weakest accuracy + short 256-token limit. Ollama all-minilm:latest = :22m = this (L6-v2); :33m = L12-v2",
            "source_url": "https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2",
            "confidence": "medium"
        },
        {
            "model_id": "nomic-ai/nomic-embed-text-v1.5",
            "model_name": "nomic-embed-text-v1.5",
            "family": "Nomic Embed v1.5 (nomic-bert-2048, BERT-based)",
            "developer": "Nomic AI",
            "params": "137M",
            "embedding_dimensions": 768,
            "matryoshka_dims": "512, 256, 128, 64",
            "max_context_tokens": 8192,
            "mteb_score": 62.280000000000001136868377216160297393798828125,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean = 62.28 (full 768-dim)",
            "approx_vram_gb": 0.299999999999999988897769753748434595763683319091796875,
            "ollama_tag": "nomic-embed-text",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": false,
            "lang_count": "1 (English)",
            "best_for": "Default general-purpose English RAG on modest hardware; long 8K docs + MRL dimension truncation. NB: Ollama shows a 2K default num_ctx but the model max is 8192 — raise num_ctx",
            "source_url": "https://huggingface.co/nomic-ai/nomic-embed-text-v1.5",
            "confidence": "high"
        },
        {
            "model_id": "ibm-granite/granite-embedding-278m-multilingual",
            "model_name": "granite-embedding-278m-multilingual",
            "family": "IBM Granite Embedding (encoder-only, XLM-RoBERTa-like; contrastive + distillation + model-merging)",
            "developer": "IBM",
            "params": "278M",
            "embedding_dimensions": 768,
            "matryoshka_dims": null,
            "max_context_tokens": 512,
            "mteb_score": 48.2000000000000028421709430404007434844970703125,
            "mteb_benchmark": "MTEB(eng) Retrieval mean = 48.2; MIRACL 18-lang mean = 58.3 (IBM Granite Embedding paper) — retrieval-only, NOT full MTEB means",
            "approx_vram_gb": 0.59999999999999997779553950749686919152736663818359375,
            "ollama_tag": "granite-embedding:278m",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "12 (en, ar, cs, de, es, fr, it, ja, ko, nl, pt, zh)",
            "best_for": "Lightweight, truly-open (Apache-2.0) multilingual retrieval; tiny+fast, but only 12 languages and 512-token window. MUST pull granite-embedding:278m — bare/:latest/:30m is the 30M English-only model",
            "source_url": "https://huggingface.co/ibm-granite/granite-embedding-278m-multilingual",
            "confidence": "high"
        },
        {
            "model_id": "sentence-transformers/paraphrase-multilingual-mpnet-base-v2",
            "model_name": "paraphrase-multilingual-mpnet-base-v2",
            "family": "Sentence-Transformers / SBERT (paraphrase, multilingual MPNet base; teacher-student distilled)",
            "developer": "sentence-transformers",
            "params": "278M",
            "embedding_dimensions": 768,
            "matryoshka_dims": null,
            "max_context_tokens": 512,
            "mteb_score": null,
            "mteb_benchmark": "Not on modern MTEB/MMTEB leaderboards — a legacy STS/paraphrase model predating MTEB-era retrieval benchmarking (expect materially weaker retrieval than E5/BGE/Granite)",
            "approx_vram_gb": 0.59999999999999997779553950749686919152736663818359375,
            "ollama_tag": "paraphrase-multilingual",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "50",
            "best_for": "Tiny, CPU-friendly multilingual semantic similarity / clustering / dedup; pull-and-go on Ollama (paraphrase-multilingual:278m), but weak for precision RAG. Model max 512 tokens though ST truncates to 128 by default",
            "source_url": "https://huggingface.co/sentence-transformers/paraphrase-multilingual-mpnet-base-v2",
            "confidence": "medium"
        },
        {
            "model_id": "mixedbread-ai/mxbai-embed-large-v1",
            "model_name": "mxbai-embed-large-v1",
            "family": "mxbai / BERT-large (mixedbread.ai)",
            "developer": "mixedbread.ai",
            "params": "335M",
            "embedding_dimensions": 1024,
            "matryoshka_dims": "512, 256 (MRL; supports arbitrary truncate_dim + binary/int8 quant)",
            "max_context_tokens": 512,
            "mteb_score": 64.68000000000000682121026329696178436279296875,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean = 64.68",
            "approx_vram_gb": 0.6999999999999999555910790149937383830547332763671875,
            "ollama_tag": "mxbai-embed-large",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": false,
            "lang_count": "1 (English)",
            "best_for": "Strong English retrieval; was SOTA for BERT-large size (Mar 2024); needs a retrieval query prompt prefix ('Represent this sentence for searching relevant passages:')",
            "source_url": "https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1",
            "confidence": "high"
        },
        {
            "model_id": "BAAI/bge-large-en-v1.5",
            "model_name": "bge-large-en-v1.5",
            "family": "BGE v1.5 Large (BERT-large)",
            "developer": "BAAI",
            "params": "335M",
            "embedding_dimensions": 1024,
            "matryoshka_dims": null,
            "max_context_tokens": 512,
            "mteb_score": 64.2300000000000039790393202565610408782958984375,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean = 64.23",
            "approx_vram_gb": 0.6999999999999999555910790149937383830547332763671875,
            "ollama_tag": "bge-large",
            "ollama_verified": true,
            "license": "MIT",
            "license_commercial": true,
            "multilingual": false,
            "lang_count": "1 (English)",
            "best_for": "Proven, permissive (MIT, commercial-OK) English RAG baseline; short 512-token chunks; prepend the retrieval instruction to queries. Ollama bge-large:335m = this model",
            "source_url": "https://huggingface.co/BAAI/bge-large-en-v1.5",
            "confidence": "high"
        },
        {
            "model_id": "nomic-ai/nomic-embed-text-v2-moe",
            "model_name": "nomic-embed-text-v2-moe",
            "family": "Nomic Embed v2 (Mixture-of-Experts, 8 experts / top-2 routing)",
            "developer": "Nomic AI",
            "params": "475M total / 305M active",
            "embedding_dimensions": 768,
            "matryoshka_dims": "256 (768->256)",
            "max_context_tokens": 512,
            "mteb_score": 65.7999999999999971578290569595992565155029296875,
            "mteb_benchmark": "MIRACL nDCG@10 mean = 65.80 (multilingual); BEIR = 52.86 (English) — model-card retrieval numbers, NOT an MTEB mean",
            "approx_vram_gb": 1,
            "ollama_tag": "nomic-embed-text-v2-moe",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "~100 (trained on 1.6B+ pairs)",
            "best_for": "Multilingual RAG wanting big-model retrieval quality at small active-param (305M) inference cost; first general-purpose MoE embedder",
            "source_url": "https://huggingface.co/nomic-ai/nomic-embed-text-v2-moe",
            "confidence": "high"
        },
        {
            "model_id": "intfloat/multilingual-e5-large-instruct",
            "model_name": "multilingual-e5-large-instruct",
            "family": "E5 (multilingual, instruction-tuned; XLM-RoBERTa-large backbone)",
            "developer": "Microsoft (intfloat)",
            "params": "560M",
            "embedding_dimensions": 1024,
            "matryoshka_dims": null,
            "max_context_tokens": 512,
            "mteb_score": 63.2000000000000028421709430404007434844970703125,
            "mteb_benchmark": "MMTEB multilingual mean ~63.2 (132 tasks) — ranked #1 among open models in the MMTEB paper",
            "approx_vram_gb": 1.100000000000000088817841970012523233890533447265625,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "MIT",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "~94 (100 inherited from XLM-RoBERTa)",
            "best_for": "Top open MMTEB-multilingual retrieval at 560M; needs a query instruction prefix for peak quality; 512-token window. Only community (not official) Ollama uploads",
            "source_url": "https://huggingface.co/intfloat/multilingual-e5-large-instruct",
            "confidence": "high"
        },
        {
            "model_id": "intfloat/multilingual-e5-large",
            "model_name": "multilingual-e5-large",
            "family": "E5 (multilingual; XLM-RoBERTa-large backbone)",
            "developer": "Microsoft (intfloat)",
            "params": "560M",
            "embedding_dimensions": 1024,
            "matryoshka_dims": null,
            "max_context_tokens": 512,
            "mteb_score": 58.60000000000000142108547152020037174224853515625,
            "mteb_benchmark": "MMTEB multilingual mean ~58.6 (132 tasks) — ranked 4th in the MMTEB paper",
            "approx_vram_gb": 1.100000000000000088817841970012523233890533447265625,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "MIT",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "~94 (100 inherited from XLM-RoBERTa)",
            "best_for": "Proven permissive multilingual RAG workhorse; no instructions, just query:/passage: prefixes; the safe MIT default — but 512-token window and only community (not official) Ollama uploads",
            "source_url": "https://huggingface.co/intfloat/multilingual-e5-large",
            "confidence": "high"
        },
        {
            "model_id": "BAAI/bge-m3",
            "model_name": "bge-m3",
            "family": "BGE-M3 (M3-Embedding, XLM-RoBERTa-large backbone; multi-functionality/linguality/granularity)",
            "developer": "BAAI",
            "params": "568M",
            "embedding_dimensions": 1024,
            "matryoshka_dims": null,
            "max_context_tokens": 8192,
            "mteb_score": 59.56000000000000227373675443232059478759765625,
            "mteb_benchmark": "No single MTEB scalar on card — headline is MIRACL(18-lang)+MKQA(25-lang) SOTA. Community-reported MMTEB multilingual mean ~59.56. Uniquely does dense + sparse + ColBERT multi-vector in one model",
            "approx_vram_gb": 1.1999999999999999555910790149937383830547332763671875,
            "ollama_tag": "bge-m3",
            "ollama_verified": true,
            "license": "MIT",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "100+",
            "best_for": "The self-host multi-vector specialist: ONE model doing dense + sparse (lexical) + native ColBERT-style late-interaction, 8K context — ideal for hybrid RAG + rerank. Ollama tag bge-m3:567m",
            "source_url": "https://huggingface.co/BAAI/bge-m3",
            "confidence": "high"
        },
        {
            "model_id": "jinaai/jina-embeddings-v3",
            "model_name": "jina-embeddings-v3",
            "family": "Jina Embeddings v3 (XLM-RoBERTa backbone + task-specific LoRA adapters)",
            "developer": "Jina AI",
            "params": "570M",
            "embedding_dimensions": 1024,
            "matryoshka_dims": "32-1024",
            "max_context_tokens": 8192,
            "mteb_score": 65.5199999999999960209606797434389591217041015625,
            "mteb_benchmark": "MTEB English mean ~65.52 (jina-reported, ~56-task original MTEB) — NOT multilingual MMTEB",
            "approx_vram_gb": 1.1999999999999999555910790149937383830547332763671875,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "CC-BY-NC-4.0",
            "license_commercial": false,
            "multilingual": true,
            "lang_count": "89 trained / ~30+ high-quality",
            "best_for": "Excellent all-round multilingual dense retrieval: 8K context, task-LoRA + Matryoshka to 32d — BUT CC-BY-NC-4.0 blocks commercial use (contact Jina); late-interaction is a separate model (jina-colbert-v2)",
            "source_url": "https://huggingface.co/jinaai/jina-embeddings-v3",
            "confidence": "high"
        },
        {
            "model_id": "Snowflake/snowflake-arctic-embed-l-v2.0",
            "model_name": "snowflake-arctic-embed-l-v2.0",
            "family": "Arctic Embed 2.0 Large (XLM-RoBERTa-large backbone)",
            "developer": "Snowflake",
            "params": "568M total (303M non-embedding)",
            "embedding_dimensions": 1024,
            "matryoshka_dims": "256 (MRL + quantization-aware; ~128 bytes/vector)",
            "max_context_tokens": 8192,
            "mteb_score": 55.60000000000000142108547152020037174224853515625,
            "mteb_benchmark": "MTEB(eng) Retrieval nDCG@10 ~55.6 (retrieval-only, NOT a full MTEB mean); multilingual MIRACL nDCG@10 ~64.9",
            "approx_vram_gb": 1.1999999999999999555910790149937383830547332763671875,
            "ollama_tag": "snowflake-arctic-embed2",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "74",
            "best_for": "Multilingual long-context RAG that must not sacrifice English retrieval; compressible (MRL+quant) vectors. Ollama tag snowflake-arctic-embed2:568m",
            "source_url": "https://huggingface.co/Snowflake/snowflake-arctic-embed-l-v2.0",
            "confidence": "medium"
        },
        {
            "model_id": "Qwen/Qwen3-Embedding-0.6B",
            "model_name": "Qwen3-Embedding-0.6B",
            "family": "Qwen3-Embedding (Alibaba Qwen)",
            "developer": "Alibaba (Qwen)",
            "params": "0.6B",
            "embedding_dimensions": 1024,
            "matryoshka_dims": "32-1024 (MRL, user-defined output dim)",
            "max_context_tokens": 32768,
            "mteb_score": 70.7000000000000028421709430404007434844970703125,
            "mteb_benchmark": "MTEB(eng, v2) mean = 70.70 (card table); MMTEB multilingual mean = 64.33",
            "approx_vram_gb": 1.5,
            "ollama_tag": "qwen3-embedding:0.6b",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "100+",
            "best_for": "Best quality-per-VRAM small embedder — 70.7 MTEB(eng,v2) at ~1.5GB fp16, 32K context, Ollama-native; strong CPU/edge RAG",
            "source_url": "https://huggingface.co/Qwen/Qwen3-Embedding-0.6B",
            "confidence": "high"
        },
        {
            "model_id": "Alibaba-NLP/gte-large-en-v1.5",
            "model_name": "gte-large-en-v1.5",
            "family": "GTE (Alibaba-NLP, BERT-based)",
            "developer": "Alibaba (Tongyi)",
            "params": "434M",
            "embedding_dimensions": 1024,
            "matryoshka_dims": null,
            "max_context_tokens": 8192,
            "mteb_score": 65.3900000000000005684341886080801486968994140625,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean = 65.39 (SOTA within size class at release)",
            "approx_vram_gb": 1.5,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": false,
            "lang_count": "1 (English)",
            "best_for": "Best small-footprint 8K-context English encoder — cheap/fast to serve at ~1.5GB; ideal high-throughput RAG default. Not in the official Ollama library",
            "source_url": "https://huggingface.co/Alibaba-NLP/gte-large-en-v1.5",
            "confidence": "high"
        },
        {
            "model_id": "NovaSearch/stella_en_1.5B_v5",
            "model_name": "stella_en_1.5B_v5",
            "family": "Stella (NovaSearch, formerly dunzhang)",
            "developer": "NovaSearch",
            "params": "1.5B",
            "embedding_dimensions": 1024,
            "matryoshka_dims": "512, 768, 1024, 2048, 4096, 6144, 8192 (separate trained projection heads; 1024 recommended)",
            "max_context_tokens": 512,
            "mteb_score": 69.43000000000000682121026329696178436279296875,
            "mteb_benchmark": "MTEB(eng, v2) mean ~69.43 (leaderboard; NOT on card). On the OLDER 56-task board it was ~71.19 — different benchmark, don't mix",
            "approx_vram_gb": 3.5,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "MIT",
            "license_commercial": true,
            "multilingual": false,
            "lang_count": "1 (English)",
            "best_for": "Efficient English RAG with flexible output dims — near-7B quality at ~1.5B/3.5GB; but only 512-token inputs (chunk small); no official Ollama tag",
            "source_url": "https://huggingface.co/NovaSearch/stella_en_1.5B_v5",
            "confidence": "medium"
        },
        {
            "model_id": "Qwen/Qwen3-Embedding-4B",
            "model_name": "Qwen3-Embedding-4B",
            "family": "Qwen3-Embedding (Alibaba Qwen)",
            "developer": "Alibaba (Qwen)",
            "params": "4B",
            "embedding_dimensions": 2560,
            "matryoshka_dims": "32-2560 (MRL, user-defined output dim)",
            "max_context_tokens": 32768,
            "mteb_score": 74.599999999999994315658113919198513031005859375,
            "mteb_benchmark": "MTEB(eng, v2) mean = 74.60 (card table); MMTEB multilingual mean = 69.45",
            "approx_vram_gb": 9,
            "ollama_tag": "qwen3-embedding:4b",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "100+",
            "best_for": "Mid-tier multilingual RAG when 0.6B underfits but a 7-8B model won't fit the GPU; ~9GB fp16",
            "source_url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B",
            "confidence": "high"
        },
        {
            "model_id": "Alibaba-NLP/gte-Qwen2-7B-instruct",
            "model_name": "gte-Qwen2-7B-instruct",
            "family": "GTE (Alibaba-NLP, Qwen2-7B backbone)",
            "developer": "Alibaba (Tongyi)",
            "params": "7B",
            "embedding_dimensions": 3584,
            "matryoshka_dims": null,
            "max_context_tokens": 32768,
            "mteb_score": 70.2399999999999948840923025272786617279052734375,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean = 70.24 (Jun 16 2024); C-MTEB (Chinese, 35 tasks) = 72.05 — not comparable to v2 boards",
            "approx_vram_gb": 15,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "EN/ZH focus; Qwen2 base is broadly multilingual",
            "best_for": "Strong bilingual EN/ZH retrieval, 32K context, commercial-friendly Apache-2.0 (2024-era SOTA); no official Ollama tag",
            "source_url": "https://huggingface.co/Alibaba-NLP/gte-Qwen2-7B-instruct",
            "confidence": "high"
        },
        {
            "model_id": "intfloat/e5-mistral-7b-instruct",
            "model_name": "e5-mistral-7b-instruct",
            "family": "E5 (intfloat / Microsoft, Mistral-7B backbone)",
            "developer": "Microsoft (intfloat)",
            "params": "7B",
            "embedding_dimensions": 4096,
            "matryoshka_dims": null,
            "max_context_tokens": 4096,
            "mteb_score": 66.6299999999999954525264911353588104248046875,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean = 66.63 (arXiv:2401.00368; NOT stated on HF card — from the paper)",
            "approx_vram_gb": 15,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "MIT",
            "license_commercial": true,
            "multilingual": false,
            "lang_count": "English-focused (some latent multilingual from Mistral; authors recommend multilingual-e5-large for non-EN)",
            "best_for": "The original LLM-as-embedder; proven MIT-licensed English 7B retriever, but now superseded by Qwen3/gte-Qwen2. Max input 4096 tokens (not 32K)",
            "source_url": "https://huggingface.co/intfloat/e5-mistral-7b-instruct",
            "confidence": "medium"
        },
        {
            "model_id": "nvidia/NV-Embed-v2",
            "model_name": "NV-Embed-v2",
            "family": "NV-Embed (NVIDIA, Mistral-7B backbone)",
            "developer": "NVIDIA",
            "params": "7.85B",
            "embedding_dimensions": 4096,
            "matryoshka_dims": null,
            "max_context_tokens": 32768,
            "mteb_score": 72.31000000000000227373675443232059478759765625,
            "mteb_benchmark": "MTEB(eng) ORIGINAL 56-task mean = 72.31 (Aug 2024, #1 at launch) — NOT comparable to Qwen3's MTEB(eng,v2)",
            "approx_vram_gb": 16,
            "ollama_tag": null,
            "ollama_verified": false,
            "license": "CC-BY-NC-4.0",
            "license_commercial": false,
            "multilingual": false,
            "lang_count": "1 (English)",
            "best_for": "Highest raw score on the original 56-task English MTEB — but CC-BY-NC-4.0 explicitly forbids commercial use; research/eval only, NOT for commercial RAG",
            "source_url": "https://huggingface.co/nvidia/NV-Embed-v2",
            "confidence": "high"
        },
        {
            "model_id": "Qwen/Qwen3-Embedding-8B",
            "model_name": "Qwen3-Embedding-8B",
            "family": "Qwen3-Embedding (Alibaba Qwen)",
            "developer": "Alibaba (Qwen)",
            "params": "8B",
            "embedding_dimensions": 4096,
            "matryoshka_dims": "32-4096 (MRL, user-defined output dim)",
            "max_context_tokens": 32768,
            "mteb_score": 75.219999999999998863131622783839702606201171875,
            "mteb_benchmark": "MTEB(eng, v2) mean = 75.22 (card table); MMTEB multilingual mean = 70.58 (#1) — do NOT conflate; both distinct from the older 56-task board",
            "approx_vram_gb": 17,
            "ollama_tag": "qwen3-embedding:8b",
            "ollama_verified": true,
            "license": "Apache-2.0",
            "license_commercial": true,
            "multilingual": true,
            "lang_count": "100+",
            "best_for": "Top open-weight retrieval embedder in 2026 (#1 MMTEB), Apache-2.0, native Ollama pull; the default when ~17GB VRAM is available",
            "source_url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B",
            "confidence": "high"
        }
    ]
}