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

mteb/leaderboard/app.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import itertools
 import json
 import logging
@@ -5,15 +7,14 @@ import tempfile
 import time
 import warnings
 from pathlib import Path
-from typing import Literal, get_args
+from typing import TYPE_CHECKING, Literal, get_args
 from urllib.parse import urlencode
 
 import cachetools
 import gradio as gr
-import pandas as pd
+import pandas as pd  # noqa: TC002 # gradio tries to validate typehints
 
 import mteb
-from mteb import BenchmarkResults
 from mteb.benchmarks.benchmark import RtebBenchmark
 from mteb.cache import ResultCache
 from mteb.leaderboard.benchmark_selector import (
@@ -31,6 +32,9 @@ from mteb.leaderboard.table import (
 from mteb.leaderboard.text_segments import ACKNOWLEDGEMENT, FAQ
 from mteb.models.model_meta import MODEL_TYPES
 
+if TYPE_CHECKING:
+    from mteb import BenchmarkResults
+
 logger = logging.getLogger(__name__)
 
 

mteb/leaderboard/table.py

@@ -1,3 +1,7 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
 import gradio as gr
 import matplotlib.pyplot as plt
 import numpy as np
@@ -5,8 +9,9 @@ import pandas as pd
 from matplotlib.colors import LinearSegmentedColormap
 from pandas.api.types import is_numeric_dtype
 
-from mteb.benchmarks.benchmark import Benchmark
-from mteb.results.benchmark_results import BenchmarkResults
+if TYPE_CHECKING:
+    from mteb.benchmarks.benchmark import Benchmark
+    from mteb.results.benchmark_results import BenchmarkResults
 
 
 def _borda_count(scores: pd.Series) -> pd.Series:

mteb/load_results.py

@@ -1,13 +1,19 @@
+from __future__ import annotations
+
 import json
 import logging
 import sys
-from collections.abc import Iterable, Sequence
-from pathlib import Path
+from typing import TYPE_CHECKING
 
 from mteb.abstasks.abstask import AbsTask
 from mteb.models.model_meta import ModelMeta
 from mteb.results import BenchmarkResults, ModelResult, TaskResult
-from mteb.types import ModelName, Revision
+
+if TYPE_CHECKING:
+    from collections.abc import Iterable, Sequence
+    from pathlib import Path
+
+    from mteb.types import ModelName, Revision
 
 if sys.version_info >= (3, 13):
     from warnings import deprecated

mteb/models/abs_encoder.py

@@ -1,14 +1,12 @@
+from __future__ import annotations
+
 import logging
 import warnings
 from abc import ABC, abstractmethod
-from collections.abc import Callable, Sequence
-from typing import Any, Literal, cast, get_args, overload
-
-from torch.utils.data import DataLoader
-from typing_extensions import Unpack
+from typing import TYPE_CHECKING, Any, Literal, cast, get_args, overload
 
 import mteb
-from mteb.abstasks.task_metadata import TaskMetadata, TaskType
+from mteb.abstasks.task_metadata import TaskType
 from mteb.similarity_functions import (
     cos_sim,
     dot_score,
@@ -18,13 +16,25 @@ from mteb.similarity_functions import (
     pairwise_max_sim,
 )
 from mteb.types import (
-    Array,
-    BatchedInput,
-    EncodeKwargs,
     PromptType,
 )
 
-from .model_meta import ModelMeta, ScoringFunction
+from .model_meta import ScoringFunction
+
+if TYPE_CHECKING:
+    from collections.abc import Callable, Sequence
+
+    from torch.utils.data import DataLoader
+    from typing_extensions import Unpack
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import (
+        Array,
+        BatchedInput,
+        EncodeKwargs,
+    )
+
+    from .model_meta import ModelMeta
 
 logger = logging.getLogger(__name__)
 
@@ -314,7 +324,7 @@ class AbsEncoder(ABC):
             ):
                 arr = self.model.similarity(embeddings1, embeddings2)
                 # We assume that the model returns an Array-like object:
-                arr = cast(Array, arr)
+                arr = cast("Array", arr)
                 return arr
             return cos_sim(embeddings1, embeddings2)
         if self.mteb_model_meta.similarity_fn_name is ScoringFunction.COSINE:
@@ -352,7 +362,7 @@ class AbsEncoder(ABC):
             ):
                 arr = self.model.similarity_pairwise(embeddings1, embeddings2)
                 # We assume that the model returns an Array-like object:
-                arr = cast(Array, arr)
+                arr = cast("Array", arr)
                 return arr
             return pairwise_cos_sim(embeddings1, embeddings2)
         if self.mteb_model_meta.similarity_fn_name is ScoringFunction.COSINE:

mteb/models/cache_wrappers/cache_backend_protocol.py

@@ -1,9 +1,11 @@
 from __future__ import annotations
 
-from pathlib import Path
-from typing import Any, Protocol, runtime_checkable
+from typing import TYPE_CHECKING, Any, Protocol, runtime_checkable
 
-import numpy as np
+if TYPE_CHECKING:
+    from pathlib import Path
+
+    import numpy as np
 
 
 @runtime_checkable

mteb/models/cache_wrappers/cache_backends/_hash_utils.py

@@ -1,6 +1,12 @@
+from __future__ import annotations
+
 import hashlib
-from collections.abc import Mapping
-from typing import Any
+from typing import TYPE_CHECKING, Any
+
+if TYPE_CHECKING:
+    from collections.abc import Mapping
+
+    from PIL import Image
 
 
 def _hash_item(item: Mapping[str, Any]) -> str:
@@ -10,8 +16,6 @@ def _hash_item(item: Mapping[str, Any]) -> str:
         item_hash = hashlib.sha256(item_text.encode()).hexdigest()
 
     if "image" in item:
-        from PIL import Image
-
         image: Image.Image = item["image"]
         item_hash += hashlib.sha256(image.tobytes()).hexdigest()
 

mteb/models/cache_wrappers/cache_backends/faiss_cache.py

@@ -1,16 +1,22 @@
+from __future__ import annotations
+
 import json
 import logging
 import warnings
 from pathlib import Path
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import numpy as np
 
 from mteb._requires_package import requires_package
-from mteb.types import BatchedInput
 
 from ._hash_utils import _hash_item
 
+if TYPE_CHECKING:
+    import faiss
+
+    from mteb.types import BatchedInput
+
 logger = logging.getLogger(__name__)
 
 
@@ -24,7 +30,6 @@ class FaissCache:
             "FAISS-based vector cache",
             install_instruction="pip install mteb[faiss-cpu]",
         )
-        import faiss
 
         self.directory = Path(directory)
         self.directory.mkdir(parents=True, exist_ok=True)

mteb/models/cache_wrappers/cache_wrapper.py

@@ -1,21 +1,26 @@
+from __future__ import annotations
+
 import logging
 from pathlib import Path
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import numpy as np
 import torch
 from datasets import Dataset
-from torch.utils.data import DataLoader
 
 from mteb._create_dataloaders import create_dataloader
-from mteb.abstasks.task_metadata import TaskMetadata
-from mteb.models.cache_wrappers.cache_backend_protocol import (
-    CacheBackendProtocol,
-)
 from mteb.models.cache_wrappers.cache_backends.numpy_cache import NumpyCache
-from mteb.models.model_meta import ModelMeta
-from mteb.models.models_protocols import EncoderProtocol
-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.models.cache_wrappers.cache_backend_protocol import (
+        CacheBackendProtocol,
+    )
+    from mteb.models.model_meta import ModelMeta
+    from mteb.models.models_protocols import EncoderProtocol
+    from mteb.types import Array, BatchedInput, PromptType
 
 logger = logging.getLogger(__name__)
 

mteb/models/get_model_meta.py

@@ -1,15 +1,22 @@
+from __future__ import annotations
+
 import difflib
 import logging
-from collections.abc import Iterable
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
-from mteb.abstasks import AbsTask
 from mteb.models import (
     ModelMeta,
-    MTEBModels,
 )
 from mteb.models.model_implementations import MODEL_REGISTRY
 
+if TYPE_CHECKING:
+    from collections.abc import Iterable
+
+    from mteb.abstasks import AbsTask
+    from mteb.models import (
+        MTEBModels,
+    )
+
 logger = logging.getLogger(__name__)
 
 

mteb/models/instruct_wrapper.py

@@ -1,16 +1,24 @@
+from __future__ import annotations
+
 import logging
-from collections.abc import Callable
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import torch
-from torch.utils.data import DataLoader
 
 from mteb._requires_package import requires_package
-from mteb.abstasks.task_metadata import TaskMetadata
-from mteb.types import Array, BatchedInput, PromptType
+from mteb.types import PromptType
 
 from .abs_encoder import AbsEncoder
 
+if TYPE_CHECKING:
+    from collections.abc import Callable
+
+    from torch.utils.data import DataLoader
+
+    from mteb.abstasks.task_metadata import TaskMetadata
+    from mteb.types import Array, BatchedInput
+
+
 logger = logging.getLogger(__name__)
 
 

mteb/models/model_implementations/align_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
 
 
 class ALIGNModel(AbsEncoder):
@@ -111,6 +116,7 @@ align_base = ModelMeta(
     release_date="2023-02-24",
     modalities=["image", "text"],
     n_parameters=176_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=671,
     max_tokens=64,
     embed_dim=768,

mteb/models/model_implementations/amazon_models.py

@@ -8,6 +8,7 @@ amazon_titan_text_embeddings_v2 = ModelMeta(
     release_date="2024-04-30",
     languages=["eng-Latn"],
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     max_tokens=None,
     embed_dim=None,

mteb/models/model_implementations/andersborges.py

@@ -12,6 +12,7 @@ model2vecdk = ModelMeta(
     revision="cb576c78dcc1b729e4612645f61db59929d69e61",
     release_date="2025-11-21",
     n_parameters=48042496,
+    n_embedding_parameters=None,
     memory_usage_mb=183,
     max_tokens=np.inf,
     embed_dim=256,
@@ -43,6 +44,7 @@ model2vecdk_stem = ModelMeta(
     revision="cb576c78dcc1b729e4612645f61db59929d69e61",
     release_date="2025-11-21",
     n_parameters=48578560,
+    n_embedding_parameters=None,
     memory_usage_mb=185,
     max_tokens=np.inf,
     embed_dim=256,

mteb/models/model_implementations/ara_models.py

@@ -10,6 +10,7 @@ arabic_triplet_matryoshka = ModelMeta(
     revision="ed357f222f0b6ea6670d2c9b5a1cb93950d34200",
     release_date="2024-07-28",
     n_parameters=135_000_000,
+    n_embedding_parameters=49_152_000,
     memory_usage_mb=516,
     embed_dim=768,
     license="apache-2.0",

mteb/models/model_implementations/arctic_models.py

@@ -147,6 +147,7 @@ arctic_embed_xs = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     n_parameters=22_600_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=86,
     max_tokens=512,
     embed_dim=384,
@@ -173,6 +174,7 @@ arctic_embed_s = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     n_parameters=32_200_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=127,
     max_tokens=512,
     embed_dim=384,
@@ -199,6 +201,7 @@ arctic_embed_m = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     n_parameters=109_000_000,
+    n_embedding_parameters=23_440_896,
     memory_usage_mb=415,
     max_tokens=512,
     embed_dim=768,
@@ -225,6 +228,7 @@ arctic_embed_m_long = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     n_parameters=137_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=522,
     max_tokens=2048,
     embed_dim=768,
@@ -250,6 +254,7 @@ arctic_embed_l = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     n_parameters=335_000_000,
+    n_embedding_parameters=31_254_528,
     memory_usage_mb=1274,
     max_tokens=512,
     embed_dim=1024,
@@ -280,6 +285,7 @@ arctic_embed_m_v1_5 = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors", "GGUF"],
     n_parameters=109_000_000,
+    n_embedding_parameters=23_440_896,
     memory_usage_mb=415,
     max_tokens=512,
     embed_dim=768,
@@ -306,6 +312,7 @@ arctic_embed_m_v2_0 = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     n_parameters=305_000_000,
+    n_embedding_parameters=None,
     memory_usage_mb=1165,
     max_tokens=8192,
     embed_dim=768,
@@ -331,6 +338,7 @@ arctic_embed_l_v2_0 = ModelMeta(
     open_weights=True,
     framework=["Sentence Transformers", "PyTorch", "ONNX", "safetensors"],
     n_parameters=568_000_000,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=2166,
     max_tokens=8192,
     embed_dim=1024,

mteb/models/model_implementations/b1ade_models.py

@@ -16,6 +16,7 @@ b1ade_embed = ModelMeta(
     open_weights=True,
     release_date="2025-03-10",
     n_parameters=335_000_000,
+    n_embedding_parameters=31_254_528,
     memory_usage_mb=1278,
     embed_dim=1024,
     license="mit",

mteb/models/model_implementations/bedrock_models.py

@@ -1,20 +1,30 @@
+from __future__ import annotations
+
 import json
 import logging
 import re
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import numpy as np
-from torch.utils.data import DataLoader
 from tqdm.auto 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_meta import ModelMeta, ScoringFunction
-from mteb.types import Array, BatchedInput, PromptType
 
-from .cohere_models import model_prompts as cohere_model_prompts
-from .cohere_models import supported_languages as cohere_supported_languages
+from .cohere_models import (
+    model_prompts as cohere_model_prompts,
+)
+from .cohere_models import (
+    supported_languages as cohere_supported_languages,
+)
+
+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__)
 
@@ -169,6 +179,7 @@ amazon_titan_embed_text_v1 = ModelMeta(
     embed_dim=1536,
     open_weights=False,
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     public_training_code=None,
     public_training_data=None,  # assumed
@@ -196,6 +207,7 @@ amazon_titan_embed_text_v2 = ModelMeta(
     embed_dim=1024,
     open_weights=False,
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     public_training_code=None,
     public_training_data=None,  # assumed
@@ -225,6 +237,7 @@ cohere_embed_english_v3 = ModelMeta(
     revision="1",
     release_date="2023-11-02",
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     public_training_code=None,
     public_training_data=None,  # assumed
@@ -253,6 +266,7 @@ cohere_embed_multilingual_v3 = ModelMeta(
     revision="1",
     release_date="2023-11-02",
     n_parameters=None,
+    n_embedding_parameters=None,
     memory_usage_mb=None,
     public_training_code=None,
     public_training_data=None,  # assumed

mteb/models/model_implementations/bge_models.py

@@ -6,7 +6,29 @@ from mteb.models.sentence_transformer_wrapper import sentence_transformers_loade
 
 from .e5_instruct import E5_MISTRAL_TRAINING_DATA
 
-model_prompts = {"query": "Represent this sentence for searching relevant passages: "}
+model_prompts = {
+    "query": "Represent this sentence for searching relevant passages: ",
+    "BrightBiologyRetrieval-query": "Represent this biology post for searching relevant passages: ",
+    "BrightEarthScienceRetrieval-query": "Represent this earth_science post for searching relevant passages: ",
+    "BrightEconomicsRetrieval-query": "Represent this economics post for searching relevant passages: ",
+    "BrightPsychologyRetrieval-query": "Represent this psychology post for searching relevant passages: ",
+    "BrightRoboticsRetrieval-query": "Represent this robotics post for searching relevant passages: ",
+    "BrightStackoverflowRetrieval-query": "Represent this stackoverflow post for searching relevant passages: ",
+    "BrightSustainableLivingRetrieval-query": "Represent this sustainable_living post for searching relevant passages: ",
+    "BrightPonyRetrieval-query": "Represent this Pony question for searching relevant passages: ",
+    "BrightLeetcodeRetrieval-query": "Represent this Coding problem for searching relevant examples: ",
+    "BrightAopsRetrieval-query": "Represent this Math problem for searching relevant examples: ",
+    "BrightTheoremQATheoremsRetrieval-query": "Represent this Math problem for searching relevant theorems: ",
+    "BrightTheoremQAQuestionsRetrieval-query": "Represent this Math problem for searching relevant examples: ",
+    "BrightBiologyLongRetrieval-query": "Represent this biology post for searching relevant documents: ",
+    "BrightEarthScienceLongRetrieval-query": "Represent this earth_science post for searching relevant documents: ",
+    "BrightEconomicsLongRetrieval-query": "Represent this economics post for searching relevant documents: ",
+    "BrightPsychologyLongRetrieval-query": "Represent this psychology post for searching relevant documents: ",
+    "BrightRoboticsLongRetrieval-query": "Represent this robotics post for searching relevant document: ",
+    "BrightStackoverflowLongRetrieval-query": "Represent this stackoverflow post for searching relevant document: ",
+    "BrightSustainableLivingLongRetrieval-query": "Represent this sustainable_living post for searching relevant documents: ",
+    "BrightPonyLongRetrieval-query": "Represent this Pony question for searching relevant documents: ",
+}
 BGE_15_CITATION = """@misc{bge_embedding,
       title={C-Pack: Packaged Resources To Advance General Chinese Embedding},
       author={Shitao Xiao and Zheng Liu and Peitian Zhang and Niklas Muennighoff},
@@ -325,6 +347,7 @@ bge_small_en_v1_5 = ModelMeta(
     revision="5c38ec7c405ec4b44b94cc5a9bb96e735b38267a",
     release_date="2023-09-12",  # initial commit of hf model.
     n_parameters=33_400_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=127,
     embed_dim=512,
     license="mit",
@@ -357,6 +380,7 @@ bge_base_en_v1_5 = ModelMeta(
     revision="a5beb1e3e68b9ab74eb54cfd186867f64f240e1a",
     release_date="2023-09-11",  # initial commit of hf model.
     n_parameters=109_000_000,
+    n_embedding_parameters=23_440_896,
     memory_usage_mb=390,
     embed_dim=768,
     license="mit",
@@ -389,6 +413,7 @@ bge_large_en_v1_5 = ModelMeta(
     revision="d4aa6901d3a41ba39fb536a557fa166f842b0e09",
     release_date="2023-09-12",  # initial commit of hf model.
     n_parameters=335_000_000,
+    n_embedding_parameters=31_254_528,
     memory_usage_mb=1242,
     embed_dim=1024,
     license="mit",
@@ -421,6 +446,7 @@ bge_small_zh = ModelMeta(
     revision="1d2363c5de6ce9ba9c890c8e23a4c72dce540ca8",
     release_date="2023-08-05",  # initial commit of hf model.
     n_parameters=33_400_000,
+    n_embedding_parameters=10_817_536,
     memory_usage_mb=127,
     embed_dim=512,
     license="mit",
@@ -448,6 +474,7 @@ bge_base_zh = ModelMeta(
     revision="0e5f83d4895db7955e4cb9ed37ab73f7ded339b6",
     release_date="2023-08-05",  # initial commit of hf model.
     n_parameters=109_000_000,
+    n_embedding_parameters=16_226_304,
     memory_usage_mb=390,
     embed_dim=768,
     license="mit",
@@ -475,6 +502,7 @@ bge_large_zh = ModelMeta(
     revision="b5d9f5c027e87b6f0b6fa4b614f8f9cdc45ce0e8",
     release_date="2023-08-02",  # initial commit of hf model.
     n_parameters=335_000_000,
+    n_embedding_parameters=21_635_072,
     memory_usage_mb=1242,
     embed_dim=1024,
     license="mit",
@@ -502,6 +530,7 @@ bge_small_en = ModelMeta(
     revision="4778d71a06863076696b03fd2777eb118712cad8",
     release_date="2023-08-05",  # initial commit of hf model.
     n_parameters=33_400_000,
+    n_embedding_parameters=11_720_448,
     memory_usage_mb=127,
     embed_dim=512,
     license="mit",
@@ -529,6 +558,7 @@ bge_base_en = ModelMeta(
     revision="b737bf5dcc6ee8bdc530531266b4804a5d77b5d8",
     release_date="2023-08-05",  # initial commit of hf model.
     n_parameters=109_000_000,
+    n_embedding_parameters=23_440_896,
     memory_usage_mb=390,
     embed_dim=768,
     license="mit",
@@ -562,6 +592,7 @@ bge_large_en = ModelMeta(
     revision="abe7d9d814b775ca171121fb03f394dc42974275",
     release_date="2023-08-05",  # initial commit of hf model.
     n_parameters=335_000_000,
+    n_embedding_parameters=31_254_528,
     memory_usage_mb=1242,
     embed_dim=1024,
     license="mit",
@@ -590,6 +621,7 @@ bge_small_zh_v1_5 = ModelMeta(
     revision="7999e1d3359715c523056ef9478215996d62a620",
     release_date="2023-09-12",  # initial commit of hf model.
     n_parameters=33_400_000,
+    n_embedding_parameters=10_817_536,
     memory_usage_mb=91,
     embed_dim=512,
     license="mit",
@@ -616,6 +648,7 @@ bge_base_zh_v1_5 = ModelMeta(
     revision="f03589ceff5aac7111bd60cfc7d497ca17ecac65",
     release_date="2023-09-11",  # initial commit of hf model.
     n_parameters=109_000_000,
+    n_embedding_parameters=16_226_304,
     memory_usage_mb=416,
     embed_dim=768,
     license="mit",
@@ -642,6 +675,7 @@ bge_large_zh_v1_5 = ModelMeta(
     revision="79e7739b6ab944e86d6171e44d24c997fc1e0116",
     release_date="2023-09-12",  # initial commit of hf model.
     n_parameters=335_000_000,
+    n_embedding_parameters=21_635_072,
     memory_usage_mb=1278,
     embed_dim=1024,
     license="mit",
@@ -665,6 +699,7 @@ bge_m3 = ModelMeta(
     revision="5617a9f61b028005a4858fdac845db406aefb181",
     release_date="2024-06-28",
     n_parameters=568_000_000,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=2167,
     embed_dim=1024,
     license="mit",
@@ -761,6 +796,7 @@ bge_multilingual_gemma2 = ModelMeta(
     revision="992e13d8984fde2c31ef8a3cb2c038aeec513b8a",
     release_date="2024-07-25",  # initial commit of hf model.
     n_parameters=int(9.24 * 1e9),
+    n_embedding_parameters=917_511_168,
     memory_usage_mb=35254,
     embed_dim=3584,  # from old C-MTEB leaderboard
     license="https://ai.google.dev/gemma/terms",
@@ -808,6 +844,7 @@ bge_en_icl = ModelMeta(
     revision="971c7e1445cc86656ca0bd85ed770b8675a40bb5",
     release_date="2024-07-25",  # initial commit of hf model.
     n_parameters=int(7.11 * 1e9),
+    n_embedding_parameters=131_084_288,
     memory_usage_mb=27125,
     embed_dim=4096,
     license="apache-2.0",
@@ -842,6 +879,7 @@ bge_m3_unsupervised = ModelMeta(
     revision="46f03bc86361cf88102b0b517b36c8259f2946b1",
     release_date="2024-01-30",  # January 30, 2024 - BGE-M3 release date
     n_parameters=568_000_000,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=2167,
     embed_dim=1024,
     license="mit",
@@ -871,6 +909,7 @@ manu__bge_m3_custom_fr = ModelMeta(
     languages=None,
     loader=sentence_transformers_loader,
     n_parameters=567754752,
+    n_embedding_parameters=256_002_048,
     memory_usage_mb=2166,
     max_tokens=8194.0,
     embed_dim=1024,