mteb 2.7.2__py3-none-any.whl → 2.7.4__py3-none-any.whl

mteb/models/model_implementations/ruri_models.py

@@ -38,6 +38,7 @@ cl_nagoya_ruri_v3_30m = ModelMeta(
     revision="24899e5de370b56d179604a007c0d727bf144504",
     release_date="2025-04-07",
     n_parameters=36_705_536,
+    n_embedding_parameters=None,
     memory_usage_mb=140,
     embed_dim=256,
     license="apache-2.0",
@@ -69,6 +70,7 @@ cl_nagoya_ruri_v3_70m = ModelMeta(
     revision="07a8b0aba47d29d2ca21f89b915c1efe2c23d1cc",
     release_date="2025-04-09",
     n_parameters=36_705_536,
+    n_embedding_parameters=None,
     memory_usage_mb=140,
     embed_dim=256,
     license="apache-2.0",
@@ -98,6 +100,7 @@ cl_nagoya_ruri_v3_130m = ModelMeta(
     revision="e3114c6ee10dbab8b4b235fbc6dcf9dd4d5ac1a6",
     release_date="2025-04-09",
     n_parameters=132_140_544,
+    n_embedding_parameters=None,
     memory_usage_mb=504,
     embed_dim=512,
     license="apache-2.0",
@@ -127,6 +130,7 @@ cl_nagoya_ruri_v3_310m = ModelMeta(
     revision="18b60fb8c2b9df296fb4212bb7d23ef94e579cd3",
     release_date="2025-04-09",
     n_parameters=314_611_968,
+    n_embedding_parameters=None,
     memory_usage_mb=1200,
     embed_dim=768,
     license="apache-2.0",
@@ -157,6 +161,7 @@ cl_nagoya_ruri_small_v2 = ModelMeta(
     revision="db18646e673b713cd0518a5bb0fefdce21e77cd9",
     release_date="2024-12-05",
     n_parameters=68_087_808,
+    n_embedding_parameters=25_165_824,
     memory_usage_mb=260,
     embed_dim=768,
     license="apache-2.0",
@@ -186,6 +191,7 @@ cl_nagoya_ruri_base_v2 = ModelMeta(
     revision="8ce03882903668a01c83ca3b8111ac025a3bc734",
     release_date="2024-12-05",
     n_parameters=111_207_168,
+    n_embedding_parameters=25_165_824,
     memory_usage_mb=424,
     embed_dim=768,
     license="apache-2.0",
@@ -215,6 +221,7 @@ cl_nagoya_ruri_large_v2 = ModelMeta(
     revision="42898ef34a5574977380ebf0dfd28cbfbd36438b",
     release_date="2024-12-06",
     n_parameters=337_441_792,
+    n_embedding_parameters=33_554_432,
     memory_usage_mb=1287,
     embed_dim=1024,
     license="apache-2.0",
@@ -245,6 +252,7 @@ cl_nagoya_ruri_small_v1 = ModelMeta(
     revision="bc56ce90cd7a979f6eb199fc52dfe700bfd94bc3",
     release_date="2024-08-28",
     n_parameters=68_087_808,
+    n_embedding_parameters=25_165_824,
     memory_usage_mb=130,
     embed_dim=768,
     license="apache-2.0",
@@ -274,6 +282,7 @@ cl_nagoya_ruri_base_v1 = ModelMeta(
     revision="1ae40b8b6c78518a499425086bab8fc16c2e4b0e",
     release_date="2024-08-28",
     n_parameters=111_207_168,
+    n_embedding_parameters=25_165_824,
     memory_usage_mb=212,
     embed_dim=768,
     license="apache-2.0",
@@ -304,6 +313,7 @@ cl_nagoya_ruri_large_v1 = ModelMeta(
     revision="a011c39b13e8bc137ee13c6bc82191ece46c414c",
     release_date="2024-08-28",
     n_parameters=337_441_792,
+    n_embedding_parameters=33_554_432,
     memory_usage_mb=644,
     embed_dim=1024,
     license="apache-2.0",

mteb/models/model_implementations/salesforce_models.py

@@ -1,12 +1,18 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
 from mteb.models.instruct_wrapper import (
     InstructSentenceTransformerModel,
     instruct_wrapper,
 )
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import PromptType
 
 from .e5_instruct import E5_MISTRAL_TRAINING_DATA
 
+if TYPE_CHECKING:
+    from mteb.types import PromptType
+
 
 def instruction_template(
     instruction: str, prompt_type: PromptType | None = None
@@ -52,6 +58,7 @@ SFR_Embedding_2_R = ModelMeta(
     revision="91762139d94ed4371a9fa31db5551272e0b83818",
     release_date="2024-06-14",  # initial commit of hf model.
     n_parameters=7_110_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=13563,
     embed_dim=4096,
     license="cc-by-nc-4.0",
@@ -90,6 +97,7 @@ SFR_Embedding_Code_2B_R = ModelMeta(
     revision="c73d8631a005876ed5abde34db514b1fb6566973",
     release_date="2025-01-17",  # initial commit of hf model.
     n_parameters=2_610_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=4986,
     embed_dim=2304,
     license="cc-by-nc-4.0",
@@ -128,6 +136,7 @@ SFR_Embedding_Mistral = ModelMeta(
     revision="938c560d1c236aa563b2dbdf084f28ab28bccb11",
     release_date="2024-01-24",  # initial commit of hf model.
     n_parameters=7_110_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=13563,
     embed_dim=4096,
     license="cc-by-nc-4.0",

mteb/models/model_implementations/samilpwc_models.py

@@ -51,6 +51,7 @@ samilpwc_expr = ModelMeta(
     revision="33358978be40f36491045f9c2a359d38c3f50047",
     release_date="2025-08-12",
     n_parameters=560_000_000,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=2136,
     embed_dim=1024,
     license="apache-2.0",

mteb/models/model_implementations/sarashina_embedding_models.py

@@ -124,6 +124,7 @@ sbintuitions_sarashina_embedding_v2_1b = ModelMeta(
     revision="1f3408afaa7b617e3445d891310a9c26dd0c68a5",
     release_date="2025-07-30",
     n_parameters=1_224_038_144,
+    n_embedding_parameters=183_500_800,
     memory_usage_mb=4669,
     embed_dim=1792,
     license="https://huggingface.co/sbintuitions/sarashina-embedding-v2-1b/blob/main/LICENSE",
@@ -150,6 +151,7 @@ sbintuitions_sarashina_embedding_v1_1b = ModelMeta(
     revision="d060fcd8984075071e7fad81baff035cbb3b6c7e",
     release_date="2024-11-22",
     n_parameters=1_224_038_144,
+    n_embedding_parameters=183_500_800,
     memory_usage_mb=4669,
     embed_dim=1792,
     license="https://huggingface.co/sbintuitions/sarashina-embedding-v1-1b/blob/main/LICENSE",

mteb/models/model_implementations/searchmap_models.py

@@ -27,6 +27,7 @@ searchmap_preview = ModelMeta(
     use_instructions=True,
     release_date="2025-03-05",
     n_parameters=435_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1660,
     embed_dim=4096,
     license="mit",

mteb/models/model_implementations/seed_1_6_embedding_models.py

@@ -13,16 +13,18 @@ import torch
 from torch.utils.data import DataLoader
 
 from mteb._requires_package import requires_package
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_implementations.bge_models import bge_chinese_training_data
 from mteb.models.model_implementations.nvidia_models import nvidia_training_datasets
 from mteb.models.model_meta import ModelMeta
-from mteb.types import Array, BatchedInput, PromptType
+from mteb.types import PromptType
 
 if TYPE_CHECKING:
     from PIL import Image
 
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput
+
 
 logger = logging.getLogger(__name__)
 
@@ -429,6 +431,7 @@ seed_embedding = ModelMeta(
     embed_dim=2048,
     open_weights=False,
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     license=None,
     reference="https://seed1-6-embedding.github.io/",

mteb/models/model_implementations/seed_1_6_embedding_models_1215.py

@@ -15,15 +15,18 @@ from torch.utils.data import DataLoader
 from tqdm import tqdm
 
 from mteb._requires_package import requires_package
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_implementations.bge_models import bge_chinese_training_data
 from mteb.models.model_implementations.nvidia_models import nvidia_training_datasets
 from mteb.models.model_meta import ModelMeta
-from mteb.types import Array, BatchedInput, PromptType
+from mteb.types import PromptType
 
 if TYPE_CHECKING:
     from PIL import Image
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput
 
 
 logger = logging.getLogger(__name__)
@@ -613,6 +616,7 @@ seed_embedding = ModelMeta(
     embed_dim=2048,
     open_weights=False,
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     license=None,
     reference="https://console.volcengine.com/ark/region:ark+cn-beijing/model/detail?Id=doubao-embedding-vision",

mteb/models/model_implementations/seed_models.py

@@ -9,7 +9,7 @@ from tqdm.auto import tqdm
 from mteb._requires_package import requires_package
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta
-from mteb.models.models_protocols import PromptType
+from mteb.types import PromptType
 
 from .bge_models import bge_chinese_training_data
 from .nvidia_models import nvidia_training_datasets
@@ -253,6 +253,7 @@ seed_embedding = ModelMeta(
     embed_dim=2048,
     open_weights=False,
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     license=None,
     reference="https://seed1-5-embedding.github.io/",

mteb/models/model_implementations/sentence_transformers_models.py

@@ -121,6 +121,7 @@ all_minilm_l6_v2 = ModelMeta(
     revision="8b3219a92973c328a8e22fadcfa821b5dc75636a",
     release_date="2021-08-30",
     n_parameters=22_700_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=87,
     embed_dim=384,
     license="apache-2.0",
@@ -152,6 +153,7 @@ all_minilm_l12_v2 = ModelMeta(
     revision="364dd28d28dcd3359b537f3cf1f5348ba679da62",
     release_date="2021-08-30",
     n_parameters=33_400_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=127,
     embed_dim=384,
     license="apache-2.0",
@@ -183,6 +185,7 @@ paraphrase_multilingual_minilm_l12_v2 = ModelMeta(
     revision="bf3bf13ab40c3157080a7ab344c831b9ad18b5eb",
     release_date="2019-11-01",  # release date of paper
     n_parameters=118_000_000,
+    n_embedding_parameters=96_014_208,
     memory_usage_mb=449,
     embed_dim=768,
     license="apache-2.0",
@@ -214,6 +217,7 @@ paraphrase_multilingual_mpnet_base_v2 = ModelMeta(
     revision="79f2382ceacceacdf38563d7c5d16b9ff8d725d6",
     release_date="2019-11-01",  # release date of paper
     n_parameters=278_000_000,
+    n_embedding_parameters=192_001_536,
     memory_usage_mb=1061,
     embed_dim=768,
     license="apache-2.0",
@@ -256,6 +260,7 @@ labse = ModelMeta(
     revision="e34fab64a3011d2176c99545a93d5cbddc9a91b7",
     release_date="2019-11-01",  # release date of paper
     n_parameters=471_000_000,
+    n_embedding_parameters=384_885_504,
     memory_usage_mb=1796,
     embed_dim=768,
     license="apache-2.0",
@@ -294,6 +299,7 @@ multi_qa_minilm_l6_cos_v1 = ModelMeta(
     revision="b207367332321f8e44f96e224ef15bc607f4dbf0",
     release_date="2021-08-30",
     n_parameters=22_700_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=87,
     embed_dim=384,
     license="apache-2.0",
@@ -325,6 +331,7 @@ all_mpnet_base_v2 = ModelMeta(
     revision="9a3225965996d404b775526de6dbfe85d3368642",
     release_date="2021-08-30",
     n_parameters=109_000_000,
+    n_embedding_parameters=23_444_736,
     memory_usage_mb=418,
     embed_dim=768,
     license="apache-2.0",
@@ -435,6 +442,7 @@ static_similarity_mrl_multilingual_v1 = ModelMeta(
     revision="7264ea07c5365a11d7e6d87dbb6195889a13054f",
     release_date="2025-01-15",
     n_parameters=108_420_096,
+    n_embedding_parameters=None,
     memory_usage_mb=413,
     embed_dim=1024,
     license="apache-2.0",
@@ -468,6 +476,7 @@ contriever = ModelMeta(
     revision="abe8c1493371369031bcb1e02acb754cf4e162fa",
     release_date="2022-06-25",  # release date of model on HF
     n_parameters=150_000_000,
+    n_embedding_parameters=23_440_896,
     memory_usage_mb=572,
     embed_dim=768,
     license=None,
@@ -498,6 +507,7 @@ microllama_text_embedding = ModelMeta(
     revision="98f70f14cdf12d7ea217ed2fd4e808b0195f1e7e",
     release_date="2024-11-10",
     n_parameters=272_000_000,
+    n_embedding_parameters=32_769_024,
     memory_usage_mb=1037,
     embed_dim=1024,
     license="apache-2.0",
@@ -544,6 +554,7 @@ sentence_t5_base = ModelMeta(
     revision="50c53e206f8b01c9621484a3c0aafce4e55efebf",
     release_date="2022-02-09",
     n_parameters=110_000_000,
+    n_embedding_parameters=24_674_304,
     memory_usage_mb=209,
     embed_dim=768,
     license="apache-2.0",
@@ -567,6 +578,7 @@ sentence_t5_large = ModelMeta(
     revision="1fc08ea477205aa54a3e5b13f0971ae16b86410a",
     release_date="2022-02-09",
     n_parameters=335_000_000,
+    n_embedding_parameters=32_899_072,
     memory_usage_mb=639,
     embed_dim=768,
     license="apache-2.0",
@@ -590,6 +602,7 @@ sentence_t5_xl = ModelMeta(
     revision="2965d31b368fb14117688e0bde77cbd720e91f53",
     release_date="2024-03-27",
     n_parameters=3_000_000_000,
+    n_embedding_parameters=32_899_072,
     memory_usage_mb=2367,
     embed_dim=768,
     license="apache-2.0",
@@ -613,6 +626,7 @@ sentence_t5_xxl = ModelMeta(
     revision="4d122282ba80e807e9e6eb8c358269e92796365d",
     release_date="2024-03-27",
     n_parameters=11_000_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=9279,
     embed_dim=768,
     license="apache-2.0",
@@ -646,6 +660,7 @@ gtr_t5_large = ModelMeta(
     revision="a2c8ac47f998531948d4cbe32a0b577a7037a5e3",
     release_date="2022-02-09",
     n_parameters=335_000_000,
+    n_embedding_parameters=32_899_072,
     memory_usage_mb=639,
     embed_dim=768,
     license="apache-2.0",
@@ -681,6 +696,7 @@ gtr_t5_xl = ModelMeta(
     revision="23a8d667a1ad2578af181ce762867003c498d1bf",
     release_date="2022-02-09",
     n_parameters=1_240_000_000,
+    n_embedding_parameters=32_899_072,
     memory_usage_mb=2367,
     embed_dim=768,
     license="apache-2.0",
@@ -715,6 +731,7 @@ gtr_t5_xxl = ModelMeta(
     revision="73f2a9156a3dcc2194dfdb2bf201cd7d17e17884",
     release_date="2022-02-09",
     n_parameters=4_860_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=9279,
     embed_dim=768,
     license="apache-2.0",
@@ -750,6 +767,7 @@ gtr_t5_base = ModelMeta(
     revision="7027e9594267928589816394bdd295273ddc0739",
     release_date="2022-02-09",
     n_parameters=110_000_000,
+    n_embedding_parameters=24_674_304,
     memory_usage_mb=209,
     embed_dim=768,
     license="apache-2.0",

mteb/models/model_implementations/shuu_model.py

@@ -10,6 +10,7 @@ codemodernbert_crow_meta = ModelMeta(
     revision="044a7a4b552f86e284817234c336bccf16f895ce",
     release_date="2025-04-21",
     n_parameters=151668480,
+    n_embedding_parameters=None,
     memory_usage_mb=607,
     embed_dim=768,
     license="apache-2.0",

mteb/models/model_implementations/siglip_models.py

@@ -1,13 +1,18 @@
-from typing import Any
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
 
 import torch
-from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
 
 SIGLIP_CITATION = """@misc{zhai2023sigmoid,
       title={Sigmoid Loss for Language Image Pre-Training},
@@ -131,6 +136,7 @@ siglip_so400m_patch14_224 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=877_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=3347,
     max_tokens=16,
     embed_dim=1152,
@@ -155,6 +161,7 @@ siglip_so400m_patch14_384 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=878_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=3349,
     max_tokens=64,
     embed_dim=1152,
@@ -179,6 +186,7 @@ siglip_so400m_patch16_256_i18n = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=1_130_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=4306,
     max_tokens=64,
     embed_dim=1152,
@@ -203,6 +211,7 @@ siglip_base_patch16_256_multilingual = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=371_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1414,
     max_tokens=64,
     embed_dim=768,
@@ -227,6 +236,7 @@ siglip_base_patch16_256 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=203_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=775,
     max_tokens=64,
     embed_dim=768,
@@ -251,6 +261,7 @@ siglip_base_patch16_512 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=204_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=777,
     max_tokens=64,
     embed_dim=768,
@@ -275,6 +286,7 @@ siglip_base_patch16_384 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=203_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=776,
     max_tokens=64,
     embed_dim=768,
@@ -299,6 +311,7 @@ siglip_base_patch16_224 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=203_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=775,
     max_tokens=64,
     embed_dim=768,
@@ -323,6 +336,7 @@ siglip_large_patch16_256 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=652_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=2488,
     max_tokens=64,
     embed_dim=1024,
@@ -347,6 +361,7 @@ siglip_large_patch16_384 = ModelMeta(
     release_date="2024-01-08",
     modalities=["image", "text"],
     n_parameters=652_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=2489,
     max_tokens=64,
     embed_dim=1024,

mteb/models/model_implementations/slm_models.py

@@ -13,24 +13,27 @@ Based on:
 from __future__ import annotations
 
 import logging
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import torch
-from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
 from mteb._requires_package import (
     requires_image_dependencies,
     requires_package,
 )
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.abs_encoder import AbsEncoder
 from mteb.models.model_implementations.colpali_models import (
     COLPALI_CITATION,
     COLPALI_TRAINING_DATA,
 )
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
 
 logger = logging.getLogger(__name__)
 

mteb/models/model_implementations/sonar_models.py

@@ -224,7 +224,8 @@ sonar = ModelMeta(
     use_instructions=False,  # it does take a language code as input
     revision="a551c586dcf4a49c8fd847de369412d556a7f2f2",
     release_date="2021-05-21",
-    n_parameters=None,  # it is really multiple models so not sure how to calculate this
+    n_parameters=None,
+    n_embedding_parameters=None,  # it is really multiple models so not sure how to calculate this
     max_tokens=512,  # https://github.com/facebookresearch/SONAR/blob/549d287466443bd8720f938047882630c1c5c3f7/sonar/models/sonar_text/builder.py#L139
     embed_dim=1024,
     license="mit",

mteb/models/model_implementations/spartan8806_atles_champion.py

@@ -12,6 +12,7 @@ spartan8806_atles_champion_embedding = ModelMeta(
     revision="d4c74d7000bbd25f3597fc0f2dcde59ef1386e8f",
     release_date="2025-11-15",
     n_parameters=110_000_000,
+    n_embedding_parameters=23_444_736,
     memory_usage_mb=420,
     max_tokens=512,
     embed_dim=768,

mteb/models/model_implementations/stella_models.py

@@ -66,6 +66,7 @@ stella_en_400m = ModelMeta(
     revision="1bb50bc7bb726810eac2140e62155b88b0df198f",
     release_date="2024-07-12",
     n_parameters=435_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1660,
     max_tokens=8192,
     embed_dim=4096,
@@ -101,6 +102,7 @@ stella_en_1_5b = ModelMeta(
     revision="d03be74b361d4eb24f42a2fe5bd2e29917df4604",
     release_date="2024-07-12",
     n_parameters=1_540_000_000,
+    n_embedding_parameters=232_928_256,
     memory_usage_mb=5887,
     max_tokens=131072,
     embed_dim=8960,
@@ -130,6 +132,7 @@ stella_large_zh_v3_1792d = ModelMeta(
     revision="d5d39eb8cd11c80a63df53314e59997074469f09",
     release_date="2024-02-17",
     n_parameters=None,
+    n_embedding_parameters=21_635_072,
     memory_usage_mb=None,  # can't see on model card
     embed_dim=1792,
     license="not specified",
@@ -157,6 +160,7 @@ stella_base_zh_v3_1792d = ModelMeta(
     revision="82254892a0fba125aa2abf3a4800d2dd12821343",
     release_date="2024-02-17",
     n_parameters=None,
+    n_embedding_parameters=16_226_304,
     memory_usage_mb=None,  # can't see on model card
     embed_dim=1792,
     license="mit",
@@ -185,6 +189,7 @@ stella_mrl_large_zh_v3_5_1792d = ModelMeta(
     revision="17bb1c32a93a8fc5f6fc9e91d5ea86da99983cfe",
     release_date="2024-02-27",
     n_parameters=int(326 * 1e6),
+    n_embedding_parameters=21_635_072,
     memory_usage_mb=1242,
     embed_dim=1792,
     license="mit",
@@ -209,6 +214,7 @@ zpoint_large_embedding_zh = ModelMeta(
     revision="b1075144f440ab4409c05622c1179130ebd57d03",
     release_date="2024-06-04",
     n_parameters=int(326 * 1e6),
+    n_embedding_parameters=21_635_072,
     memory_usage_mb=1242,
     embed_dim=1792,
     license="mit",

mteb/models/model_implementations/tarka_models.py

@@ -327,6 +327,7 @@ tarka_embedding_150m_v1 = ModelMeta(
     revision="b0ffecc4ef0d873e517507ed080e43b88b2704b9",
     release_date="2025-11-04",
     n_parameters=155_714_304,
+    n_embedding_parameters=None,
     embed_dim=768,
     max_tokens=2048,
     license="gemma",
@@ -361,6 +362,7 @@ tarka_embedding_350m_v1 = ModelMeta(
     revision="a850d6a329145474727424fed6b12b62096b8ba3",
     release_date="2025-11-11",
     n_parameters=354_483_968,
+    n_embedding_parameters=None,
     memory_usage_mb=676,
     embed_dim=1024,
     max_tokens=128000,

mteb/models/model_implementations/text2vec_models.py

@@ -22,6 +22,7 @@ text2vec_base_chinese = ModelMeta(
     revision="183bb99aa7af74355fb58d16edf8c13ae7c5433e",
     release_date="2022-01-23",
     n_parameters=int(102 * 1e6),
+    n_embedding_parameters=16_226_304,
     embed_dim=768,
     license="apache-2.0",
     max_tokens=512,
@@ -51,6 +52,7 @@ text2vec_base_chinese_paraphrase = ModelMeta(
     revision="e90c150a9c7fb55a67712a766d6820c55fb83cdd",
     release_date="2023-06-19",
     n_parameters=118 * 1e6,
+    n_embedding_parameters=30_720_000,
     memory_usage_mb=450,
     embed_dim=768,
     license="apache-2.0",
@@ -95,6 +97,7 @@ text2vec_base_multilingual = ModelMeta(
     # So probably best not to.
     loader=sentence_transformers_loader,
     n_parameters=117654272,
+    n_embedding_parameters=96_014_208,
     memory_usage_mb=449,
     embed_dim=384,
     license="apache-2.0",

mteb/models/model_implementations/ua_sentence_models.py

@@ -8,6 +8,7 @@ xlm_roberta_ua_distilled = ModelMeta(
     model_type=["dense"],
     loader=sentence_transformers_loader,
     n_parameters=278_000_000,
+    n_embedding_parameters=192_001_536,
     memory_usage_mb=1061,
     max_tokens=512,
     embed_dim=768,

mteb/models/model_implementations/uae_models.py

@@ -1,13 +1,18 @@
+from __future__ import annotations
+
 import logging
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import torch
-from torch.utils.data import DataLoader
 
-from mteb.abstasks.task_metadata import TaskMetadata
 from mteb.models.model_meta import ModelMeta, ScoringFunction
 from mteb.models.sentence_transformer_wrapper import SentenceTransformerEncoderWrapper
-from mteb.types import Array, BatchedInput, PromptType
+
+if TYPE_CHECKING:
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput, PromptType
 
 logger = logging.getLogger(__name__)
 
@@ -67,6 +72,7 @@ uae_large_v1 = ModelMeta(
     revision="369c368f70f16a613f19f5598d4f12d9f44235d4",
     release_date="2023-12-04",  # initial commit of hf model.
     n_parameters=int(335 * 1e6),
+    n_embedding_parameters=31_254_528,
     memory_usage_mb=1278,
     max_tokens=512,
     embed_dim=1024,

mteb/models/model_implementations/vdr_models.py

@@ -1,6 +1,12 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
 from mteb.models.instruct_wrapper import InstructSentenceTransformerModel
 from mteb.models.model_meta import ModelMeta, ScoringFunction
-from mteb.types import PromptType
+
+if TYPE_CHECKING:
+    from mteb.types import PromptType
 
 
 def instruction_template(
@@ -32,6 +38,7 @@ vdr_2b_multi_v1 = ModelMeta(
     release_date="2024-01-08",
     modalities=["text"],  # TODO: integrate with image
     n_parameters=2_000_000_000,
+    n_embedding_parameters=233_373_696,
     memory_usage_mb=4213,
     max_tokens=32768,
     embed_dim=1536,