Model comparison
The models you picked, side by side. A comparison covers one primitive
(e.g. /encode) at a time.
No models selected yet. Add some from the catalog.
drives latency, throughput & cost across every table on this page
| Model | Size | Quality | Latency | Throughput | Cost $/1M | |
|---|---|---|---|---|---|---|
| Alibaba-NLP/gte-Qwen2-7B-instruct Long contextDense | 7.6B | 0.4055ndcg@10 | 846 ms | 3.5K tok/s | $0.063 | |
| Qwen/Qwen3-Embedding-8B Long contextDense | 7.6B | 0.4140ndcg@10 | — | — | — | |
| GritLM/GritLM-7B Dense | 7.2B | 0.3452ndcg@10 | 2.1 s | 1.4K tok/s | $0.157 | |
| Linq-AI-Research/Linq-Embed-Mistral Long contextDense | 7.1B | 0.4192ndcg@10 | 818 ms | 2.9K tok/s | $0.075 | |
| Salesforce/SFR-Embedding-2_R Long contextDense | 7.1B | 0.4142ndcg@10 | 865 ms | 3.0K tok/s | $0.075 | |
| Salesforce/SFR-Embedding-Mistral Dense | 7.1B | 0.4085ndcg@10 | 888 ms | 3.0K tok/s | $0.075 | |
| intfloat/e5-mistral-7b-instruct Dense | 7.1B | 0.3890ndcg@10 | 915 ms | 3.0K tok/s | $0.074 | |
| vidore/colqwen2.5-v0.2 MultimodalMulti-vector | 7.0B | — | 1.9 s | 7.6 mpix/s | — | |
| nvidia/nemotron-colembed-vl-4b-v2 MultimodalMultilingualLong contextMulti-vector | 4.8B | — | — | — | — | |
| TomoroAI/tomoro-colqwen3-embed-4b MultimodalMultilingualLong contextMulti-vector | 4.4B | — | — | — | — | |
| nvidia/llama-nemoretriever-colembed-3b-v1 MultimodalMultilingualLong contextMulti-vector | 4.4B | — | 6.1 s | 0.7 img/s | — | |
| Qwen/Qwen3-Embedding-4B Long contextDense | 4.0B | 0.4121ndcg@10 | 464 ms | 5.7K tok/s | $0.039 | |
| vidore/colpali-v1.3-hf MultimodalMulti-vector | 3.0B | — | 582 ms | 23.0 mpix/s | — | |
| Qwen/Qwen3-VL-Embedding-2B MultimodalLong contextDense | 2.1B | 0.3689ndcg@10 | 36 ms | 494 tok/s | $0.450 | |
| Alibaba-NLP/gte-Qwen2-1.5B-instruct Long contextDense | 1.8B | 0.3937ndcg@10 | 261 ms | 12.3K tok/s | $0.018 | |
| NovaSearch/stella_en_1.5B_v5 Dense | 1.5B | 0.4215ndcg@10 | 258 ms | 12.8K tok/s | $0.017 | |
| laion/CLIP-ViT-H-14-laion2B-s32B-b79K MultimodalDense | 986M | — | 353 ms | 438 tok/s | $0.508 | |
| google/siglip-so400m-patch14-384 MultimodalDense | 878M | — | 554 ms | 290 tok/s | $0.766 | |
| google/siglip-so400m-patch14-224 MultimodalDense | 877M | — | 284 ms | 456 tok/s | $0.487 | |
| Qwen/Qwen3-Embedding-0.6B Long contextDense | 596M | 0.3689ndcg@10 | 157 ms | 20.6K tok/s | $0.011 | |
| tencent/R3-embedding-0.6b Dense | 596M | — | 197 ms | 24.5K tok/s | $0.0091 | |
| BAAI/bge-m3 Long contextDenseSparseMulti-vector | 568M | 0.3147ndcg@10 | 93 ms | 33.2K tok/s | $0.0067 | |
| Snowflake/snowflake-arctic-embed-l-v2.0 MultilingualLong contextDense | 568M | 0.3519ndcg@10 | 107 ms | 42.9K tok/s | $0.0052 | |
| intfloat/multilingual-e5-large MultilingualDense | 560M | 0.3213ndcg@10 | 109 ms | 29.8K tok/s | $0.0074 | |
| intfloat/multilingual-e5-large-instruct MultilingualDense | 560M | 0.3521ndcg@10 | 107 ms | 29.4K tok/s | $0.0076 | |
| jinaai/jina-colbert-v2 MultilingualLong contextMulti-vector | 559M | 0.3566ndcg@10 | 106 ms | 28.5K tok/s | $0.0078 | |
| nomic-ai/nomic-embed-text-v2-moe MultilingualDense | 475M | 0.3461ndcg@10 | 150 ms | 13.0K tok/s | $0.017 | |
| NovaSearch/stella_en_400M_v5 Dense | 435M | 0.4125ndcg@10 | 116 ms | 27.1K tok/s | $0.0082 | |
| openai/clip-vit-large-patch14 MultimodalDense | 428M | — | 228 ms | 977 tok/s | $0.227 | |
| google/siglip2-base-patch16-224 MultimodalDense | 375M | — | 159 ms | 699 tok/s | $0.318 | |
| mixedbread-ai/mxbai-colbert-large-v1 Multi-vector | 335M | 0.3467ndcg@10 | 75 ms | 43.3K tok/s | $0.0051 | |
| intfloat/e5-large-v2 Dense | 335M | 0.3315ndcg@10 | 87 ms | 33.2K tok/s | $0.0067 | |
| mixedbread-ai/mxbai-embed-large-v1 Dense | 335M | 0.3865ndcg@10 | — | — | — | |
| naver/v-splade-quality MultimodalSparse | 330M | — | — | — | — | |
| lightonai/mLateOn MultilingualLong contextMulti-vector | 307M | 0.3776ndcg@10 | — | — | — | |
| Alibaba-NLP/gte-multilingual-base MultilingualLong contextDense | 305M | 0.3690ndcg@10 | 57 ms | 55.1K tok/s | $0.0040 | |
| Snowflake/snowflake-arctic-embed-m-v2.0 MultilingualLong contextDense | 305M | 0.2489ndcg@10 | — | — | — | |
| google/embeddinggemma-300m Dense | 303M | 0.3928ndcg@10 | 87 ms | 27.2K tok/s | $0.0082 | |
| Marqo/marqo-ecommerce-embeddings-B MultimodalDense | 203M | — | — | — | — | |
| Marqo/marqo-fashionSigLIP MultimodalDense | 203M | — | — | — | — | |
| laion/CLIP-ViT-B-32-laion2B-s34B-b79K MultimodalDense | 151M | — | 219 ms | 1.0K tok/s | $0.218 | |
| openai/clip-vit-base-patch32 MultimodalDense | 151M | — | 234 ms | 958 tok/s | $0.232 | |
| lightonai/GTE-ModernColBERT-v1 Long contextMulti-vector | 149M | 0.3618ndcg@10 | 99 ms | 28.0K tok/s | $0.0079 | |
| lightonai/Reason-ModernColBERT Long contextMulti-vector | 149M | 0.3817ndcg@10 | 82 ms | 33.0K tok/s | $0.0067 | |
| Alibaba-NLP/gte-modernbert-base Long contextDense | 149M | 0.3664ndcg@10 | — | — | — | |
| ibm-granite/granite-embedding-english-r2 Long contextDense | 149M | 0.3450ndcg@10 | — | — | — | |
| nomic-ai/modernbert-embed-base Long contextDense | 149M | 0.3337ndcg@10 | — | — | — | |
| opensearch-project/opensearch-neural-sparse-encoding-doc-v3-gte Sparse | 137M | 0.3524ndcg@10 | 94 ms | 34.2K tok/s | $0.0065 | |
| opensearch-project/opensearch-neural-sparse-encoding-v1 Sparse | 133M | 0.3600ndcg@10 | 69 ms | 48.7K tok/s | $0.0046 | |
| intfloat/multilingual-e5-small MultilingualDense | 118M | 0.3101ndcg@10 | 28 ms | 85.0K tok/s | $0.0026 | |
| naver/splade-cocondenser-selfdistil Sparse | 110M | 0.3403ndcg@10 | 72 ms | 40.0K tok/s | $0.0056 | |
| naver/splade-v3 Sparse | 110M | 0.3632ndcg@10 | 84 ms | 29.6K tok/s | $0.0075 | |
| prithivida/Splade_PP_en_v2 Sparse | 110M | 0.3306ndcg@10 | 55 ms | 57.5K tok/s | $0.0039 | |
| colbert-ir/colbertv2.0 Multi-vector | 110M | 0.3431ndcg@10 | 66 ms | 43.0K tok/s | $0.0052 | |
| intfloat/e5-base-v2 Dense | 109M | 0.3116ndcg@10 | 58 ms | 53.2K tok/s | $0.0042 | |
| ibm-granite/granite-embedding-97m-multilingual-r2 MultilingualLong contextDense | 97M | 0.2920ndcg@10 | 30 ms | 82.6K tok/s | $0.0027 | |
| opensearch-project/opensearch-neural-sparse-encoding-doc-v2-distill Sparse | 67M | 0.3396ndcg@10 | 63 ms | 49.1K tok/s | $0.0045 | |
| opensearch-project/opensearch-neural-sparse-encoding-doc-v3-distill Sparse | 67M | 0.3308ndcg@10 | 61 ms | 50.1K tok/s | $0.0044 | |
| opensearch-project/opensearch-neural-sparse-encoding-v2-distill Sparse | 67M | 0.3373ndcg@10 | 63 ms | 44.2K tok/s | $0.0050 | |
| ibm-granite/granite-embedding-small-english-r2 Long contextDense | 48M | 0.3016ndcg@10 | — | — | — | |
| answerdotai/answerai-colbert-small-v1 Multi-vector | 33M | 0.3715ndcg@10 | 48 ms | 59.1K tok/s | $0.0038 | |
| Snowflake/snowflake-arctic-embed-s Dense | 33M | 0.3265ndcg@10 | 27 ms | 99.0K tok/s | $0.0022 | |
| intfloat/e5-small-v2 Dense | 33M | 0.3195ndcg@10 | 50 ms | 58.3K tok/s | $0.0038 | |
| mixedbread-ai/mxbai-edge-colbert-v0-32m Long contextMulti-vector | 32M | 0.3560ndcg@10 | 60 ms | 45.9K tok/s | $0.0048 | |
| ibm-granite/granite-embedding-30m-sparse Sparse | 30M | 0.3147ndcg@10 | 105 ms | 31.9K tok/s | $0.0070 | |
| opensearch-project/opensearch-neural-sparse-encoding-doc-v2-mini Sparse | 23M | 0.3267ndcg@10 | 55 ms | 51.1K tok/s | $0.0044 | |
| sentence-transformers/all-MiniLM-L6-v2 Dense | 23M | 0.2324ndcg@10 | 53 ms | 55.3K tok/s | $0.0040 | |
| rasyosef/splade-mini Sparse | 11M | 0.3090ndcg@10 | 56 ms | 56.3K tok/s | $0.0039 | |
| No models match. | ||||||
| Model | Size | Quality | Latency | Throughput | Cost $/1M | |
|---|---|---|---|---|---|---|
| Qwen/Qwen3-Reranker-4B Long context | 4.0B | 0.6968ndcg@10 | 339 ms | 550 tok/s | $0.404 | |
| Qwen/Qwen3-VL-Reranker-2B MultimodalLong context | 2.1B | 0.6657ndcg@10 | 173 ms | 258 tok/s | $0.861 | |
| mixedbread-ai/mxbai-rerank-large-v2 MultilingualLong context | 1.5B | 0.6914ndcg@10 | 767 ms | 1.9K tok/s | $0.118 | |
| Qwen/Qwen3-Reranker-0.6B Long context | 596M | 0.6519ndcg@10 | 81 ms | 1.6K tok/s | $0.138 | |
| tencent/R3-rerank-0.6b | 596M | — | 449 ms | 2.1K tok/s | $0.103 | |
| BAAI/bge-m3 Long context | 568M | 0.6657ndcg@10 | 56 ms | 2.9K tok/s | $0.076 | |
| BAAI/bge-reranker-v2-m3 MultilingualLong context | 568M | 0.6763ndcg@10 | 92 ms | 30.0K tok/s | $0.0074 | |
| BAAI/bge-reranker-large Multilingual | 560M | 0.6404ndcg@10 | 52 ms | 21.1K tok/s | $0.011 | |
| jinaai/jina-colbert-v2 MultilingualLong context | 559M | 0.6391ndcg@10 | 226 ms | 1.4K tok/s | $0.158 | |
| mixedbread-ai/mxbai-rerank-base-v2 MultilingualLong context | 494M | 0.6638ndcg@10 | 451 ms | 7.0K tok/s | $0.032 | |
| mixedbread-ai/mxbai-colbert-large-v1 | 335M | 0.6299ndcg@10 | 46 ms | 4.0K tok/s | $0.056 | |
| lightonai/mLateOn MultilingualLong context | 307M | 0.6627ndcg@10 | — | — | — | |
| jinaai/jina-reranker-v2-base-multilingual Multilingual | 278M | 0.6546ndcg@10 | 38 ms | 29.0K tok/s | $0.0077 | |
| BAAI/bge-reranker-base Multilingual | 278M | 0.5926ndcg@10 | 45 ms | 21.3K tok/s | $0.010 | |
| Alibaba-NLP/gte-reranker-modernbert-base Long context | 150M | 0.6701ndcg@10 | 55 ms | 11.0K tok/s | $0.020 | |
| lightonai/GTE-ModernColBERT-v1 Long context | 149M | 0.6091ndcg@10 | 313 ms | 231 tok/s | $0.961 | |
| lightonai/Reason-ModernColBERT Long context | 149M | 0.6338ndcg@10 | — | — | — | |
| colbert-ir/colbertv2.0 | 110M | 0.5910ndcg@10 | 51 ms | 3.8K tok/s | $0.058 | |
| answerdotai/answerai-colbert-small-v1 | 33M | 0.6259ndcg@10 | 122 ms | 1.7K tok/s | $0.128 | |
| cross-encoder/ms-marco-MiniLM-L-12-v2 | 33M | 0.6145ndcg@10 | 40 ms | 26.4K tok/s | $0.0084 | |
| mixedbread-ai/mxbai-edge-colbert-v0-32m Long context | 32M | 0.6106ndcg@10 | — | — | — | |
| cross-encoder/ms-marco-MiniLM-L-6-v2 | 23M | 0.6027ndcg@10 | 46 ms | 51.1K tok/s | $0.0043 | |
| No models match. | ||||||
| Model | Size | Quality | Latency | Throughput | Cost $/1M | |
|---|---|---|---|---|---|---|
| zai-org/GLM-OCR MultimodalMultilingualLong contextEntities | 1.3B | — | 16.8 s | 662 tok/s | $0.336 | |
| opendatalab/MinerU2.5-Pro-2604-1.2B MultimodalMultilingualEntities | 1.2B | — | — | — | — | |
| lightonai/LightOnOCR-2-1B MultimodalMultilingualLong contextEntities | 1.0B | — | 20.8 s | 579 tok/s | $0.384 | |
| PaddlePaddle/PaddleOCR-VL-1.5 MultimodalMultilingualLong contextEntities | 959M | — | 12.4 s | 989 tok/s | $0.225 | |
| openai/whisper-large-v3-turbo Multilingual | 809M | — | 1.0 s | — | — | |
| microsoft/Florence-2-large MultimodalText regions | 777M | — | — | — | — | |
| fastino/gliner2-large-v1 MultilingualEntities | 486M | 0.5401F1 | — | — | — | |
| numind/NuNER_Zero Entities | 449M | 0.6068F1 | 107 ms | 14.9K tok/s | $0.015 | |
| knowledgator/gliclass-large-v3.0 | 439M | — | 94 ms | 27.2K tok/s | $0.0082 | |
| knowledgator/gliclass-large-v1.0 | 438M | — | — | — | — | |
| google/owlv2-large-patch14-ensemble MultimodalBounding boxes | 438M | — | — | — | — | |
| MoritzLaurer/deberta-v3-large-zeroshot-v2.0 | 435M | — | — | — | — | |
| EmergentMethods/gliner_large_news-v2.1 Entities | 435M | 0.5527F1 | — | — | — | |
| Ihor/gliner-biomed-large-v1.0 Entities | 435M | 0.6439F1 | 108 ms | 9.9K tok/s | $0.023 | |
| gliner-community/gliner_large-v2.5 MultilingualEntities | 435M | 0.6551F1 | — | — | — | |
| jackboyla/glirel-large-v0 Relations | 435M | — | 105 ms | 7.3K tok/s | $0.030 | |
| urchade/gliner_large-v2.1 MultilingualEntities | 435M | 0.5483F1 | 175 ms | 5.9K tok/s | $0.037 | |
| urchade/gliner_multi_pii-v1 MultilingualEntities | 435M | 0.5377F1 | 66 ms | 23.3K tok/s | $0.0096 | |
| facebook/bart-large-mnli | 407M | — | — | — | — | |
| urchade/gliner_multi-v2.1 MultilingualEntities | 289M | 0.6020F1 | — | — | — | |
| mynkchaudhry/Florence-2-FT-DocVQA MultimodalText regions | 271M | — | 1.6 s | 510 tok/s | $0.436 | |
| IDEA-Research/grounding-dino-base MultimodalBounding boxes | 233M | — | 549 ms | 0.8 mpix/s | — | |
| microsoft/Florence-2-base MultimodalText regions | 232M | — | — | — | — | |
| microsoft/Florence-2-base-ft MultimodalText regions | 232M | — | — | — | — | |
| fastino/gliguard-LLMGuardrails-300M | 208M | — | 55 ms | 47.5K tok/s | $0.0047 | |
| fastino/gliner2-base-v1 Entities | 208M | 0.5194F1 | 135 ms | 11.0K tok/s | $0.020 | |
| urchade/gliner_medium-v2.1 Entities | 195M | 0.6111F1 | 107 ms | 8.9K tok/s | $0.025 | |
| knowledgator/gliclass-base-v1.0 | 186M | — | — | — | — | |
| cross-encoder/nli-deberta-v3-base | 184M | — | — | — | — | |
| MoritzLaurer/deberta-v3-base-zeroshot-v2.0 | 184M | — | — | — | — | |
| IDEA-Research/grounding-dino-tiny MultimodalBounding boxes | 172M | — | 331 ms | 0.9 mpix/s | — | |
| google/owlv2-base-patch16-ensemble MultimodalBounding boxes | 155M | — | 955 ms | 1.0 mpix/s | — | |
| gliner-community/gliner_medium-v2.5 MultilingualEntities | 150M | 0.6642F1 | — | — | — | |
| MoritzLaurer/ModernBERT-base-zeroshot-v2.0 | 150M | — | — | — | — | |
| knowledgator/gliclass-small-v1.0 | 144M | — | — | — | — | |
| naver-clova-ix/donut-base-finetuned-cord-v2 MultimodalText regions | 110M | — | 8.4 s | 757 tok/s | $0.294 | |
| naver-clova-ix/donut-base-finetuned-docvqa MultimodalText regions | 110M | — | 6.9 s | 87 tok/s | $2.56 | |
| numind/NuNER_Zero-span Entities | 110M | 0.6448F1 | — | — | — | |
| superlinked/docling-artifacts MultimodalOCR-Document | 80M | — | 13.2 s | 380 tok/s | $0.584 | |
| gliner-community/gliner_small-v2.5 MultilingualEntities | 60M | 0.6691F1 | — | — | — | |
| urchade/gliner_small-v2.1 Entities | 60M | 0.5959F1 | 83 ms | 11.7K tok/s | $0.019 | |
| knowledgator/gliner-bi-base-v2.0 Entities | null | 0.6396F1 | 133 ms | 7.1K tok/s | $0.031 | |
| knowledgator/modern-gliner-bi-base-v1.0 Entities | null | 0.6644F1 | 127 ms | 7.3K tok/s | $0.030 | |
| No models match. | ||||||
| Model | Size | Quality | Latency | Throughput | Cost $/1M | |
|---|---|---|---|---|---|---|
| Qwen/Qwen3.6-27B MultimodalLong contextTool callingConstrained outputStreamingCodeSQL | 27.0B | 0.6000acc | 1.7 s | 222 tok/s | $3.80 | |
| Qwen/Qwen3-4B-Instruct-2507 Long contextTool callingConstrained outputStreamingCodeSQL | 4.0B | 0.6033acc | 576 ms | 472 tok/s | $1.78 | |
| Qwen/Qwen3.5-4B MultimodalLong contextTool callingConstrained outputStreaming | 4.0B | — | 762 ms | 353 tok/s | $2.38 | |
| ibm-granite/granite-guardian-3.0-2b Long contextStreamingGuard | 2.5B | — | — | — | — | |
| Qwen/Qwen3-0.6B Streaming | 600M | 0.4600acc | 413 ms | 595 tok/s | $1.41 | |
| No models match. | ||||||
Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.
Compare (0)Compare →