The best embedding models in 2026
The best embedding models right now are DINOv3, Voyage rerank-3 and Harrier-OSS-v1 (270M, 0.6B, 27B), out of 23 we have reviewed. Each is ranked by our rating, with its price, our verdict and who it suits.
By ReviewRamble editors · Updated Nov 9, 2026 · How we review
How we ranked: by our rating out of 5, from reviews built on each product's own pricing, documentation and policies plus independent benchmarks where they exist, each checked on the date shown in the review. We take no affiliate commission.
- 1
DINOv3 is the vision backbone a computer-vision team should start from in 2026, provided legal is happy with Meta's licence.
Best for: Computer-vision teams that want one frozen image encoder to feed detection, segmentation, depth or retrieval heads, including on satellite and aerial imagery.
Price: Free under the DINOv3 License after a manually approved request; commercial use allowed, with military, ITAR, nuclear, espionage and reverse-engineering restrictions. No hosted API from Meta.
Read our DINOv3 review4.6/5 - 2
On Voyage's own numbers rerank-3 leads every hosted rival, and the lite model costs so little that the rate limit, not the bill, decides how much reranking you do.
Best for: Teams running RAG or search in production who want the best hosted reranking quality for the money.
Price: rerank-3 $0.05 per 1M tokens; rerank-3-lite $0.02 per 1M tokens; first 200M tokens free.
Read our Voyage rerank-3 review4.6/5 - 3
Harrier is the most accurate open multilingual text embedder on the leaderboard that matters, and its small sizes are the better story.
Best for: Teams that self-host multilingual text search and want the best open accuracy per parameter with no licence strings.
Price: Free under MIT. No hosted API from Microsoft; the models can be deployed on your own Azure compute from the Foundry catalogue, billed as normal compute.
Read our Harrier-OSS-v1 (270M, 0.6B, 27B) review4.5/5 - 4
A strong retrieval model with one trick rivals lack, a shared vector space, but its lead rests on Voyage's own tests and it may train on your data until an admin opts out.
Best for: Teams building text search or RAG who want to pay for quality at index time and run cheap or local query embeddings.
Price: $0.12 per 1M tokens; first 200M tokens free (not on Batch); Batch API 33% off.
Read our voyage-4-large review4.5/5 - 5
Rerank 5 makes reranking cheap enough to stop rationing and calibrated enough to trust a threshold, though every benchmark behind it so far is Cohere's own.
Best for: Teams running production RAG or enterprise search who want a hosted reranker with dependable score thresholds and private-deployment options.
Price: Pro $0.05 and Fast $0.02 per million tokens on Cohere's API. Model Vault dedicated instance $4.00/hour ($2,500/month) for either size.
Read our Cohere Rerank 5 review4.4/5 - 6
Nemotron 3.5 Embed is the open embedder to start with for document search, because it reads page images as well as text and costs nothing to try.
Best for: Teams building search or RAG over PDFs, slides and reports who want one open model for both text and page images.
Price: Free under OpenMDW-1.1. Both sizes are free to call on OpenRouter; Baseten, DeepInfra and FriendliAI also host them, and NVIDIA offers a NIM container.
Read our Nemotron 3.5 Embed (1B and 8B) review4.4/5 - 7
EmbeddingGemma 2 makes private, offline search across photos, voice notes and code practical on a phone, while leaving text-only retrieval quality almost exactly where version one left it.
Best for: Mobile and edge developers building private or offline search across text, code, images and audio.
Price: Free under Apache 2.0, subject to the Gemma Prohibited Use Policy. No hosted API from Google or Hugging Face inference providers yet.
Read our EmbeddingGemma 2 review4.3/5 - 8
Gemini Embedding 2 is the simplest way to search video and audio with a sentence, and an overpriced way to search plain text.
Best for: Developers building search or recommendations across mixed media such as images, recordings and video.
Price: Per 1M tokens: text $0.20, image $0.45, audio $6.50, video $12; Batch half price; free tier with limits (inputs used for product improvement).
Read our Gemini Embedding 2 review4.3/5 - 9
Jina v5's shared text-and-media vector space is the cleverest idea in small embeddings, and a non-commercial licence is the toll for anyone who wants to run it themselves.
Best for: Teams building multilingual or multimodal search on small models, especially inside Elastic.
Price: Jina API: small models $0.05 per 1M tokens, nano $0.02. Weights free for non-commercial use; commercial self-hosting needs a paid Elastic licence.
Read our Jina Embeddings v5 review4.3/5 - 10
A 0.6B reranker that beats a 4B one on BEIR, held back by a licence that turns self-hosting into a sales call.
Best for: Elastic users and teams that want a small, long-context reranker on their own hardware.
Price: Jina API $0.05 per 1M tokens. Weights free for non-commercial use; commercial self-hosting needs a paid Elastic licence.
Read our Jina Reranker v3.5 review4.3/5 - 11
The open embedder to start with when one index has to hold slides, screenshots and video alongside text, with Apache 2.0 settling the licence question.
Best for: Teams self-hosting search across images, video and mixed documents under a permissive licence.
Price: Free under Apache 2.0.
Read our Qwen3-VL-Embedding review4.3/5 - 12
Built for long, visually messy business documents and now priced in the open, though every benchmark that flatters it is Cohere's own.
Best for: Enterprises indexing long PDFs, filings and slide decks with mixed text and images.
Price: API: Pro $0.12 and Fast $0.08 per 1M text tokens; images $0.40 per 1M. Model Vault from $3/hour or $2,000/month per instance.
Read our Cohere Embed 5 review4.2/5 - 13
A complete open search stack under one licence, embedder and reranker together, though its leaderboard crown dates from June 2025.
Best for: Teams self-hosting multilingual or code search who want both embedding and reranking under one open licence.
Price: Free under Apache 2.0. Qwen Cloud's hosted text-embedding-v4, built on Qwen3, is $0.07 per 1M tokens.
Read our Qwen3-Embedding and Qwen3-Reranker review4.2/5 - 14
V-JEPA 2 is the open video encoder to beat for motion and robot planning, but it hands you features and a research codebase, not a finished product.
Best for: Robotics and video-ML researchers who need a frozen, permissively licensed video encoder for action recognition, anticipation or planning from goal images.
Price: Free. Code under MIT; weights released for commercial and research use, with Hugging Face cards listing MIT or Apache 2.0. No hosted API from Meta.
Read our V-JEPA 2 and V-JEPA 2.1 review4.2/5 - 15
Nemotron 3.5 Rerank is the open reranker for document search where the answer may be in a chart, but at 8B it is a heavy second stage for plain text.
Best for: Teams running retrieval over PDFs, slides and scanned documents who need the most precise open second stage.
Price: Free under OpenMDW-1.1. Free to call on OpenRouter for now; NVIDIA NIM available, with no public price.
Read our Nemotron 3.5 Rerank 8B review4.1/5 - 16
Granite Embedding R2 is the small embedder a legal department will sign off quickly, though Harrier's 270M edges it on accuracy.
Best for: Enterprises that need a small, multilingual embedder with clear data provenance for search over long documents.
Price: Free under Apache 2.0. The model cards list no hosted endpoint; self-host on CPU or GPU.
Read our Granite Embedding Multilingual R2 (311M and 97M) review3.9/5 - 17
Purpose-built for finding moments in video, with small vectors that keep storage cheap, though every number behind it comes from TwelveLabs.
Best for: Teams building search over large video libraries who want a managed, video-specialist embedding API and will run their own vector store.
Price: Embed API at launch, per 1M tokens: video $0.26, audio $0.065, image $0.08, document $0.162, text $0.20. Plans: Free, Plus $29/month, Pro $99/month, Enterprise custom.
Read our Marengo 3.5 review3.9/5 - 18
A sharp open reranker whose calibrated scores are worth having, now with nobody behind it but your own GPU.
Best for: Teams already running GPU inference who want a free reranker with relevance scores they can use as a cut-off.
Price: Free under Apache 2.0. ZeroEntropy's hosted API ($0.025 per 1M tokens at launch) was supported only until 4 September 2026 and now returns an unavailable error.
Read our zerank-2 review3.9/5 - 19
Still the safest small reranker to self-host on licence alone, though its benchmark lead dates from March 2025.
Best for: Teams that want a small, permissively licensed reranker to run on their own hardware.
Price: Free under Apache 2.0. Hosted only inside Mixedbread Search: semantic search $4 per 1,000 queries, $7.50 with reranking; Scale plan $20/month in credits.
Read our mxbai-rerank v2 review3.8/5 - 20
Accurate enough to be the yardstick for page-image search, and licensed so that a yardstick is all most companies can use it as.
Best for: Researchers and evaluation teams who need a strong open baseline for visual document retrieval and do not plan to ship it.
Price: Free weights for non-commercial and research use (CC BY-NC 4.0).
Read our Nemotron ColEmbed V2 review3.8/5 - 21
Rerank 4 still works, but Cohere's own Rerank 5 now beats it on every benchmark Cohere reports, which makes it a model to migrate from rather than adopt.
Best for: Teams with a tuned Rerank 4 pipeline who need time before re-validating on Rerank 5.
Price: Billed per search (one query, up to 100 documents). Vercel AI Gateway lists Pro at $2.50 and Fast at $2.00 per 1,000 searches; Cohere's page no longer shows Rerank 4 rates.
Read our Cohere Rerank 4 review3.7/5 - 22
Safe, cheap and supported by every vector database, text-embedding-3-large now trails newer retrievers by enough to justify a re-test before any new build.
Best for: Teams already on OpenAI who value one vendor, strict data terms and broad tool support over top retrieval quality.
Price: $0.13 per 1M tokens; Batch half price; text-embedding-3-small $0.02 per 1M.
Read our text-embedding-3-large review3.7/5 - 23
Still the fastest way to get semantic search working, and no longer the right model to ship.
Best for: Learners and prototypers who need a fast, universally supported English embedding model.
Price: Free under Apache 2.0.
Read our all-MiniLM-L6-v2 review3.4/5
Sources
The main official source behind each ranked review; every review lists its full set.
- DINOv3: Meta AI: DINOv3
- Voyage rerank-3: Voyage AI: rerank-3 and rerank-3-lite
- Harrier-OSS-v1 (270M, 0.6B, 27B): microsoft/harrier-oss-v1-27b model card
- voyage-4-large: Voyage AI: The Voyage 4 model family
- Cohere Rerank 5: Cohere blog: Rerank 5
- Nemotron 3.5 Embed (1B and 8B): nvidia/Nemotron-3.5-Embed-1B-BF16 model card
- EmbeddingGemma 2: EmbeddingGemma 2 model card
- Gemini Embedding 2: Gemini API docs: embeddings
- Jina Embeddings v5: Jina AI Embeddings API
- Jina Reranker v3.5: jinaai/jina-reranker-v3.5 model card
- Qwen3-VL-Embedding: Qwen3-VL-Embedding-8B model card
- Cohere Embed 5: Cohere docs: Embed models
- Qwen3-Embedding and Qwen3-Reranker: Qwen3-Embedding-0.6B model card
- V-JEPA 2 and V-JEPA 2.1: Meta AI: V-JEPA 2 research page
- Nemotron 3.5 Rerank 8B: nvidia/Nemotron-3.5-Rerank-8B-BF16 model card
- Granite Embedding Multilingual R2 (311M and 97M): ibm-granite/granite-embedding-311m-multilingual-r2 model card
- Marengo 3.5: Marengo 3.5: Multimodal Video Embedding Model (TwelveLabs)
- zerank-2: zeroentropy/zerank-2-reranker model card
- mxbai-rerank v2: mixedbread-ai/mxbai-rerank-large-v2 model card
- Nemotron ColEmbed V2: nvidia/nemotron-colembed-vl-8b-v2 model card
- Cohere Rerank 4: Cohere docs: Rerank models
- text-embedding-3-large: OpenAI docs: text-embedding-3-large
- all-MiniLM-L6-v2: all-MiniLM-L6-v2 model card
More guides
- Best large language models (LLMs)
- Best AI reasoning models
- Best AI coding models
- Best small and on-device AI models
- Best vision and multimodal AI models
- Best OCR and document AI models
- Best AI image models
- Best AI video models
- Best AI decision models
- Best text-to-speech models
- Best speech-to-text models
- Best AI music and audio models
- Best open-source LLMs
- Best American open-source LLMs
- Best Chinese open-source LLMs
- Best European open-source LLMs