mteb 2.5.1__py3-none-any.whl → 2.5.3__py3-none-any.whl

mteb/models/model_implementations/nomic_models.py

@@ -199,6 +199,7 @@ nomic_embed_v1_5 = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="nomic-ai/nomic-embed-text-v1.5",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="b0753ae76394dd36bcfb912a46018088bca48be0",
@@ -227,6 +228,7 @@ nomic_embed_v1 = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="nomic-ai/nomic-embed-text-v1",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="0759316f275aa0cb93a5b830973843ca66babcf5",
@@ -255,6 +257,7 @@ nomic_embed_v1_ablated = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="nomic-ai/nomic-embed-text-v1-ablated",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="7d948905c5d5d3874fa55a925d68e49dbf411e5f",
@@ -282,6 +285,7 @@ nomic_embed_v1_unsupervised = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="nomic-ai/nomic-embed-text-v1-unsupervised",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="b53d557b15ae63852847c222d336c1609eced93c",
@@ -309,6 +313,7 @@ nomic_modern_bert_embed = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="nomic-ai/modernbert-embed-base",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="5960f1566fb7cb1adf1eb6e816639cf4646d9b12",
@@ -447,6 +452,7 @@ nomic_embed_text_v2_moe = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="nomic-ai/nomic-embed-text-v2-moe",
+    model_type=["dense"],
     languages=m_languages,
     open_weights=True,
     revision="1066b6599d099fbb93dfcb64f9c37a7c9e503e85",

mteb/models/model_implementations/nomic_models_vision.py

@@ -168,6 +168,7 @@ nomic_embed_vision_v1_5 = ModelMeta(
         "text_model_revision": "a03db6748c80237063eb0546ac6b627eca2318cb",
     },
     name="nomic-ai/nomic-embed-vision-v1.5",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="af2246fffdab78d8458418480e4886a8e48b70a7",
     release_date="2024-06-08",

mteb/models/model_implementations/nvidia_llama_nemoretriever_colemb.py

@@ -146,6 +146,7 @@ llama_nemoretriever_colembed_1b_v1 = ModelMeta(
         trust_remote_code=True,
     ),
     name="nvidia/llama-nemoretriever-colembed-1b-v1",
+    model_type=["late-interaction"],
     languages=["eng-Latn"],
     revision="1f0fdea7f5b19532a750be109b19072d719b8177",
     release_date="2025-06-27",
@@ -172,6 +173,7 @@ llama_nemoretriever_colembed_3b_v1 = ModelMeta(
         trust_remote_code=True,
     ),
     name="nvidia/llama-nemoretriever-colembed-3b-v1",
+    model_type=["late-interaction"],
     languages=["eng-Latn"],
     revision="50c36f4d5271c6851aa08bd26d69f6e7ca8b870c",
     release_date="2025-06-27",

mteb/models/model_implementations/nvidia_models.py

@@ -111,6 +111,7 @@ NV_embed_v2 = ModelMeta(
         add_eos_token=True,
     ),
     name="nvidia/NV-Embed-v2",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="7604d305b621f14095a1aa23d351674c2859553a",
@@ -141,6 +142,7 @@ NV_embed_v1 = ModelMeta(
         add_eos_token=True,
     ),
     name="nvidia/NV-Embed-v1",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="570834afd5fef5bf3a3c2311a2b6e0a66f6f4f2c",
@@ -528,6 +530,7 @@ class LlamaEmbedNemotron(AbsEncoder):
 llama_embed_nemotron_8b = ModelMeta(
     loader=LlamaEmbedNemotron,
     name="nvidia/llama-embed-nemotron-8b",
+    model_type=["dense"],
     languages=llama_embed_nemotron_evaluated_languages,
     open_weights=True,
     revision="84a375593d27d3528beb4e104822515659e093b4",

mteb/models/model_implementations/octen_models.py

@@ -0,0 +1,195 @@
+from mteb.models.instruct_wrapper import InstructSentenceTransformerModel
+from mteb.models.model_meta import ModelMeta
+from mteb.models.models_protocols import PromptType
+
+
+def instruction_template(
+    instruction: str, prompt_type: PromptType | None = None
+) -> str:
+    if (
+        prompt_type == PromptType.document
+    ):  # to avoid this issue: https://huggingface.co/Qwen/Qwen3-Embedding-8B/discussions/21
+        return " "
+    if not instruction:
+        return ""
+    if isinstance(instruction, dict):
+        if prompt_type is None:
+            instruction = next(iter(instruction.values()))  # TODO
+        else:
+            instruction = instruction[prompt_type]
+    return f"Instruct: {instruction}\nQuery:"
+
+
+multilingual_langs = [
+    "afr-Latn",
+    "ara-Arab",
+    "aze-Latn",
+    "bel-Cyrl",
+    "bul-Cyrl",
+    "ben-Beng",
+    "cat-Latn",
+    "ceb-Latn",
+    "ces-Latn",
+    "cym-Latn",
+    "dan-Latn",
+    "deu-Latn",
+    "ell-Grek",
+    "eng-Latn",
+    "spa-Latn",
+    "est-Latn",
+    "eus-Latn",
+    "fas-Arab",
+    "fin-Latn",
+    "fra-Latn",
+    "glg-Latn",
+    "guj-Gujr",
+    "heb-Hebr",
+    "hin-Deva",
+    "hrv-Latn",
+    "hat-Latn",
+    "hun-Latn",
+    "hye-Armn",
+    "ind-Latn",
+    "isl-Latn",
+    "ita-Latn",
+    "jpn-Jpan",
+    "jav-Latn",
+    "kat-Geor",
+    "kaz-Cyrl",
+    "khm-Khmr",
+    "kan-Knda",
+    "kor-Hang",
+    "kir-Cyrl",
+    "lao-Laoo",
+    "lit-Latn",
+    "lav-Latn",
+    "mkd-Cyrl",
+    "mal-Mlym",
+    "mon-Cyrl",
+    "mar-Deva",
+    "msa-Latn",
+    "mya-Mymr",
+    "nep-Deva",
+    "nld-Latn",
+    "nor-Latn",
+    "nob-Latn",
+    "nno-Latn",
+    "pan-Guru",
+    "pol-Latn",
+    "por-Latn",
+    "que-Latn",
+    "ron-Latn",
+    "rus-Cyrl",
+    "sin-Sinh",
+    "slk-Latn",
+    "slv-Latn",
+    "swa-Latn",
+    "tam-Taml",
+    "tel-Telu",
+    "tha-Thai",
+    "tgl-Latn",
+    "tur-Latn",
+    "ukr-Cyrl",
+    "urd-Arab",
+    "vie-Latn",
+    "yor-Latn",
+    "zho-Hans",
+]
+
+OCTEN_CITATION = """@misc{octen-embedding-2025,
+  title={Octen-Embedding-8B: A Fine-tuned Multilingual Text Embedding Model},
+  author={Octen Team},
+  year={2025},
+  url={https://huggingface.co/bflhc/bflhc/Octen-Embedding-8B}
+}"""
+
+training_data = {
+    "T2Retrieval",
+    "DuRetrieval",
+    "MMarcoReranking",
+    "CMedQAv2-reranking",
+    "NQ",
+    "MSMARCO",
+    "HotpotQA",
+    "FEVER",
+    "MrTidyRetrieval",
+    "MIRACLRetrieval",
+    "CodeSearchNet",
+}
+
+# Predefined prompts for various RTEB tasks
+_PREDEFINED_PROMPTS = {
+    # ========== Open Datasets ==========
+    # Legal domain
+    "AILACasedocs": "Given a legal case scenario, retrieve the most relevant case documents",
+    "AILAStatutes": "Given a legal scenario, retrieve the most relevant statute documents",
+    "LegalQuAD": "Given a legal question, retrieve relevant legal documents that answer the question",
+    "LegalSummarization": "Given a query, retrieve relevant legal documents for summarization",
+    # Code domain
+    "AppsRetrieval": "Given a query about mobile applications, retrieve relevant app information",
+    "HumanEvalRetrieval": "Given a code problem description, retrieve relevant code examples",
+    "MBPPRetrieval": "Given a programming problem description, retrieve relevant code solutions",
+    "DS1000Retrieval": "Given a data science problem, retrieve relevant code snippets",
+    "FreshStackRetrieval": "Given a programming question, retrieve relevant Stack Overflow posts",
+    # Finance domain
+    "FinQARetrieval": "Given a financial question, retrieve relevant financial documents",
+    "FinanceBenchRetrieval": "Given a financial query, retrieve relevant financial information",
+    "HC3FinanceRetrieval": "Given a finance-related query, retrieve relevant documents",
+    # Medical domain
+    "CUREv1": "Given a medical query, retrieve relevant clinical documents",
+    "ChatDoctorRetrieval": "Given a medical question, retrieve relevant medical information",
+    # SQL domain
+    "WikiSQLRetrieval": "Given a natural language query, retrieve relevant SQL examples",
+    # Multilingual
+    "MIRACLRetrievalHardNegatives": "Given a question, retrieve Wikipedia passages that answer the question",
+    # ========== Private/Closed Datasets ==========
+    # Code domain (Private)
+    "Code1Retrieval": "Given a code problem description, retrieve relevant code examples",
+    "JapaneseCode1Retrieval": "Given a code problem description, retrieve relevant code examples",
+    # Finance domain (Private)
+    "EnglishFinance1Retrieval": "Given a financial query, retrieve relevant financial documents",
+    "EnglishFinance2Retrieval": "Given a financial query, retrieve relevant financial documents",
+    "EnglishFinance3Retrieval": "Given a financial query, retrieve relevant financial documents",
+    "EnglishFinance4Retrieval": "Given a financial query, retrieve relevant financial documents",
+    # Healthcare domain (Private)
+    "EnglishHealthcare1Retrieval": "Given a medical question, retrieve relevant medical information",
+    "GermanHealthcare1Retrieval": "Given a medical question, retrieve relevant medical information",
+    # Legal domain (Private)
+    "FrenchLegal1Retrieval": "Given a legal query, retrieve relevant legal documents",
+    "GermanLegal1Retrieval": "Given a legal query, retrieve relevant legal documents",
+    "JapaneseLegal1Retrieval": "Given a legal query, retrieve relevant legal documents",
+    # General/Multilingual (Private)
+    "French1Retrieval": "Given a query, retrieve relevant passages",
+    "German1Retrieval": "Given a query, retrieve relevant passages",
+}
+
+
+Octen_Embedding_8B = ModelMeta(
+    loader=InstructSentenceTransformerModel,
+    loader_kwargs=dict(
+        instruction_template=instruction_template,
+        apply_instruction_to_passages=True,
+        prompts_dict=_PREDEFINED_PROMPTS,
+        max_seq_length=18480,
+        model_kwargs={"torch_dtype": "bfloat16"},
+    ),
+    name="bflhc/Octen-Embedding-8B",
+    languages=multilingual_langs,
+    open_weights=True,
+    revision="2030603c2926ab005fafd824fac5911e271be21f",
+    release_date="2025-12-23",
+    n_parameters=7567295488,
+    memory_usage_mb=14433,
+    embed_dim=4096,
+    max_tokens=32768,
+    license="apache-2.0",
+    reference="https://huggingface.co/bflhc/Octen-Embedding-8B",
+    similarity_fn_name="cosine",
+    framework=["Sentence Transformers", "PyTorch"],
+    use_instructions=True,
+    public_training_code=None,
+    public_training_data=None,
+    training_datasets=training_data,
+    citation=OCTEN_CITATION,
+    adapted_from="Qwen/Qwen3-Embedding-8B",
+)

mteb/models/model_implementations/openai_models.py

@@ -167,6 +167,7 @@ class OpenAIModel(AbsEncoder):
 
 text_embedding_3_small = ModelMeta(
     name="openai/text-embedding-3-small",
+    model_type=["dense"],
     revision="3",
     release_date="2024-01-25",
     languages=None,  # supported languages not specified
@@ -191,6 +192,7 @@ text_embedding_3_small = ModelMeta(
 )
 text_embedding_3_large = ModelMeta(
     name="openai/text-embedding-3-large",
+    model_type=["dense"],
     revision="3",
     release_date="2024-01-25",
     languages=None,  # supported languages not specified
@@ -215,6 +217,7 @@ text_embedding_3_large = ModelMeta(
 )
 text_embedding_ada_002 = ModelMeta(
     name="openai/text-embedding-ada-002",
+    model_type=["dense"],
     revision="3",
     release_date="2022-12-15",
     languages=None,  # supported languages not specified
@@ -240,6 +243,7 @@ text_embedding_ada_002 = ModelMeta(
 
 text_embedding_3_small_512 = ModelMeta(
     name="openai/text-embedding-3-small (embed_dim=512)",
+    model_type=["dense"],
     revision="3",
     release_date="2024-01-25",
     languages=None,  # supported languages not specified
@@ -266,6 +270,7 @@ text_embedding_3_small_512 = ModelMeta(
 
 text_embedding_3_large_512 = ModelMeta(
     name="openai/text-embedding-3-large (embed_dim=512)",
+    model_type=["dense"],
     revision="3",
     release_date="2024-01-25",
     languages=None,  # supported languages not specified

mteb/models/model_implementations/openclip_models.py

@@ -122,6 +122,7 @@ def openclip_loader(model_name, **kwargs):
 CLIP_ViT_L_14_DataComp_XL_s13B_b90K = ModelMeta(
     loader=openclip_loader,  # type: ignore
     name="laion/CLIP-ViT-L-14-DataComp.XL-s13B-b90K",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="84c9828e63dc9a9351d1fe637c346d4c1c4db341",
     release_date="2023-04-26",
@@ -147,6 +148,7 @@ CLIP_ViT_L_14_DataComp_XL_s13B_b90K = ModelMeta(
 CLIP_ViT_B_32_DataComp_XL_s13B_b90K = ModelMeta(
     loader=openclip_loader,  # type: ignore
     name="laion/CLIP-ViT-B-32-DataComp.XL-s13B-b90K",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="f0e2ffa09cbadab3db6a261ec1ec56407ce42912",
     release_date="2023-04-26",
@@ -172,6 +174,7 @@ CLIP_ViT_B_32_DataComp_XL_s13B_b90K = ModelMeta(
 CLIP_ViT_B_16_DataComp_XL_s13B_b90K = ModelMeta(
     loader=openclip_loader,  # type: ignore
     name="laion/CLIP-ViT-B-16-DataComp.XL-s13B-b90K",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="d110532e8d4ff91c574ee60a342323f28468b287",
     release_date="2023-04-26",
@@ -197,6 +200,7 @@ CLIP_ViT_B_16_DataComp_XL_s13B_b90K = ModelMeta(
 CLIP_ViT_bigG_14_laion2B_39B_b160k = ModelMeta(
     loader=openclip_loader,  # type: ignore
     name="laion/CLIP-ViT-bigG-14-laion2B-39B-b160k",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="bc7788f151930d91b58474715fdce5524ad9a189",
     release_date="2023-01-23",
@@ -222,6 +226,7 @@ CLIP_ViT_bigG_14_laion2B_39B_b160k = ModelMeta(
 CLIP_ViT_g_14_laion2B_s34B_b88K = ModelMeta(
     loader=openclip_loader,  # type: ignore
     name="laion/CLIP-ViT-g-14-laion2B-s34B-b88K",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="15efd0f6ac0c40c0f9da7becca03c974d7012604",
     release_date="2023-03-06",
@@ -247,6 +252,7 @@ CLIP_ViT_g_14_laion2B_s34B_b88K = ModelMeta(
 CLIP_ViT_H_14_laion2B_s32B_b79K = ModelMeta(
     loader=openclip_loader,  # type: ignore
     name="laion/CLIP-ViT-H-14-laion2B-s32B-b79K",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="de081ac0a0ca8dc9d1533eed1ae884bb8ae1404b",
     release_date="2022-09-15",
@@ -272,6 +278,7 @@ CLIP_ViT_H_14_laion2B_s32B_b79K = ModelMeta(
 CLIP_ViT_L_14_laion2B_s32B_b82K = ModelMeta(
     loader=openclip_loader,  # type: ignore
     name="laion/CLIP-ViT-L-14-laion2B-s32B-b82K",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="1627032197142fbe2a7cfec626f4ced3ae60d07a",
     release_date="2022-09-15",
@@ -297,6 +304,7 @@ CLIP_ViT_L_14_laion2B_s32B_b82K = ModelMeta(
 CLIP_ViT_B_32_laion2B_s34B_b79K = ModelMeta(
     loader=openclip_loader,
     name="laion/CLIP-ViT-B-32-laion2B-s34B-b79K",
+    model_type=["dense"],
     languages=["eng-Latn"],
     revision="08f73555f1b2fb7c82058aebbd492887a94968ef",
     release_date="2022-09-15",

mteb/models/model_implementations/opensearch_neural_sparse_models.py

@@ -128,6 +128,7 @@ class SparseEncoderWrapper(AbsEncoder):
 
 opensearch_neural_sparse_encoding_doc_v3_gte = ModelMeta(
     name="opensearch-project/opensearch-neural-sparse-encoding-doc-v3-gte",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="a8abaa916125ee512a7a8f4d706d07eb0128a8e6",
@@ -153,6 +154,7 @@ opensearch_neural_sparse_encoding_doc_v3_gte = ModelMeta(
 
 opensearch_neural_sparse_encoding_doc_v3_distill = ModelMeta(
     name="opensearch-project/opensearch-neural-sparse-encoding-doc-v3-distill",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="babf71f3c48695e2e53a978208e8aba48335e3c0",
@@ -174,6 +176,7 @@ opensearch_neural_sparse_encoding_doc_v3_distill = ModelMeta(
 
 opensearch_neural_sparse_encoding_doc_v2_distill = ModelMeta(
     name="opensearch-project/opensearch-neural-sparse-encoding-doc-v2-distill",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="8921a26c78b8559d6604eb1f5c0b74c079bee38f",
@@ -196,6 +199,7 @@ opensearch_neural_sparse_encoding_doc_v2_distill = ModelMeta(
 
 opensearch_neural_sparse_encoding_doc_v2_mini = ModelMeta(
     name="opensearch-project/opensearch-neural-sparse-encoding-doc-v2-mini",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="4af867a426867dfdd744097531046f4289a32fdd",
@@ -217,6 +221,7 @@ opensearch_neural_sparse_encoding_doc_v2_mini = ModelMeta(
 
 opensearch_neural_sparse_encoding_doc_v1 = ModelMeta(
     name="opensearch-project/opensearch-neural-sparse-encoding-doc-v1",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="98cdcbd72867c547f72f2b7b7bed9cdf9f09922d",

mteb/models/model_implementations/ops_moa_models.py

@@ -22,6 +22,7 @@ class OPSWrapper(AbsEncoder):
 
 ops_moa_conan_embedding = ModelMeta(
     name="OpenSearch-AI/Ops-MoA-Conan-embedding-v1",
+    model_type=["dense"],
     revision="46dcd58753f3daa920c66f89e47086a534089350",
     release_date="2025-03-26",
     languages=["zho-Hans"],
@@ -53,6 +54,7 @@ ops_moa_conan_embedding = ModelMeta(
 
 ops_moa_yuan_embedding = ModelMeta(
     name="OpenSearch-AI/Ops-MoA-Yuan-embedding-1.0",
+    model_type=["dense"],
     revision="23712d0766417b0eb88a2513c6e212a58b543268",
     release_date="2025-03-26",
     languages=["zho-Hans"],

mteb/models/model_implementations/pawan_models.py

@@ -14,6 +14,7 @@ PAWAN_EMBD_CITATION = """@misc{medhi2025pawanembd,
 pawan_embd_68m = ModelMeta(
     loader=sentence_transformers_loader,
     name="dmedhi/PawanEmbd-68M",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="32f295145802bdbd65699ad65fd27d2a5b69a909",

mteb/models/model_implementations/piccolo_models.py

@@ -6,6 +6,7 @@ from mteb.models.sentence_transformer_wrapper import sentence_transformers_loade
 piccolo_base_zh = ModelMeta(
     loader=sentence_transformers_loader,
     name="sensenova/piccolo-base-zh",
+    model_type=["dense"],
     languages=["zho-Hans"],
     open_weights=True,
     revision="47c0a63b8f667c3482e05b2fd45577bb19252196",
@@ -29,6 +30,7 @@ piccolo_base_zh = ModelMeta(
 piccolo_large_zh_v2 = ModelMeta(
     loader=sentence_transformers_loader,
     name="sensenova/piccolo-large-zh-v2",
+    model_type=["dense"],
     languages=["zho-Hans"],
     open_weights=False,  # They "temporarily" removed it in may last year
     # "Due to certain internal company considerations"

mteb/models/model_implementations/promptriever_models.py

@@ -75,6 +75,7 @@ promptriever_llama2 = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="samaya-ai/promptriever-llama2-7b-v1",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="01c7f73d771dfac7d292323805ebc428287df4f9-30b14e3813c0fa45facfd01a594580c3fe5ecf23",  # base-peft revision
@@ -106,6 +107,7 @@ promptriever_llama3 = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="samaya-ai/promptriever-llama3.1-8b-v1",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="48d6d0fc4e02fb1269b36940650a1b7233035cbb-2ead22cfb1b0e0c519c371c63c2ab90ffc511b8a",  # base-peft revision
@@ -138,6 +140,7 @@ promptriever_llama3_instruct = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="samaya-ai/promptriever-llama3.1-8b-instruct-v1",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="5206a32e0bd3067aef1ce90f5528ade7d866253f-8b677258615625122c2eb7329292b8c402612c21",  # base-peft revision
@@ -170,6 +173,7 @@ promptriever_mistral_v1 = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="samaya-ai/promptriever-mistral-v0.1-7b-v1",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="7231864981174d9bee8c7687c24c8344414eae6b-876d63e49b6115ecb6839893a56298fadee7e8f5",  # base-peft revision

mteb/models/model_implementations/pylate_models.py

@@ -337,6 +337,7 @@ class MultiVectorModel(AbsEncoder, PylateSearchEncoder):
 colbert_v2 = ModelMeta(
     loader=MultiVectorModel,
     name="colbert-ir/colbertv2.0",
+    model_type=["late-interaction"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="c1e84128e85ef755c096a95bdb06b47793b13acf",
@@ -369,6 +370,7 @@ jina_colbert_v2 = ModelMeta(
         trust_remote_code=True,
     ),
     name="jinaai/jina-colbert-v2",
+    model_type=["late-interaction"],
     languages=[
         "ara-Arab",
         "ben-Beng",
@@ -445,6 +447,7 @@ jina_colbert_v2 = ModelMeta(
 lightonai__gte_moderncolbert_v1 = ModelMeta(
     loader=MultiVectorModel,
     name="lightonai/GTE-ModernColBERT-v1",
+    model_type=["late-interaction"],
     languages=[
         "eng-Latn",
     ],

mteb/models/model_implementations/qodo_models.py

@@ -30,6 +30,7 @@ qodo_languages = [
 Qodo_Embed_1_1_5B = ModelMeta(
     loader=sentence_transformers_loader,
     name="Qodo/Qodo-Embed-1-1.5B",
+    model_type=["dense"],
     languages=qodo_languages,
     open_weights=True,
     revision="84bbef079b32e8823ec226d4e9e92902706b0eb6",
@@ -52,6 +53,7 @@ Qodo_Embed_1_1_5B = ModelMeta(
 Qodo_Embed_1_7B = ModelMeta(
     loader=sentence_transformers_loader,
     name="Qodo/Qodo-Embed-1-7B",
+    model_type=["dense"],
     languages=qodo_languages,
     open_weights=True,
     revision="f9edd9bf7f687c0e832424058e265120f603cd81",

mteb/models/model_implementations/qtack_models.py

@@ -25,6 +25,7 @@ mini_gte_datasets = {
 mini_gte = ModelMeta(
     loader=sentence_transformers_loader,
     name="prdev/mini-gte",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="7fbe6f9b4cc42615e0747299f837ad7769025492",

mteb/models/model_implementations/qwen3_models.py

@@ -134,6 +134,7 @@ def q3e_instruct_loader(
 Qwen3_Embedding_0B6 = ModelMeta(
     loader=q3e_instruct_loader,
     name="Qwen/Qwen3-Embedding-0.6B",
+    model_type=["dense"],
     languages=multilingual_langs,
     open_weights=True,
     revision="b22da495047858cce924d27d76261e96be6febc0",  # Commit of @tomaarsen
@@ -156,6 +157,7 @@ Qwen3_Embedding_0B6 = ModelMeta(
 Qwen3_Embedding_4B = ModelMeta(
     loader=q3e_instruct_loader,
     name="Qwen/Qwen3-Embedding-4B",
+    model_type=["dense"],
     languages=multilingual_langs,
     open_weights=True,
     revision="636cd9bf47d976946cdbb2b0c3ca0cb2f8eea5ff",  # Commit of @tomaarsen
@@ -178,6 +180,7 @@ Qwen3_Embedding_4B = ModelMeta(
 Qwen3_Embedding_8B = ModelMeta(
     loader=q3e_instruct_loader,
     name="Qwen/Qwen3-Embedding-8B",
+    model_type=["dense"],
     languages=multilingual_langs,
     open_weights=True,
     revision="4e423935c619ae4df87b646a3ce949610c66241c",  # Commit of @tomaarsen

mteb/models/model_implementations/qzhou_models.py

@@ -58,6 +58,7 @@ QZhou_Embedding = ModelMeta(
         apply_instruction_to_passages=False,
     ),
     name="Kingsoft-LLM/QZhou-Embedding",
+    model_type=["dense"],
     languages=["eng-Latn", "zho-Hans"],
     open_weights=True,
     revision="f1e6c03ee3882e7b9fa5cec91217715272e433b8",
@@ -91,6 +92,7 @@ QZhou_Embedding_Zh = ModelMeta(
         apply_instruction_to_passages=False,
     ),
     name="Kingsoft-LLM/QZhou-Embedding-Zh",
+    model_type=["dense"],
     languages=["zho-Hans"],
     open_weights=True,
     revision="0321ccb126413d1e49c5ce908e802b63d35f18e2",

mteb/models/model_implementations/random_baseline.py

@@ -189,6 +189,7 @@ class RandomEncoderBaseline:
 random_encoder_baseline = ModelMeta(
     loader=RandomEncoderBaseline,  # type: ignore
     name="baseline/random-encoder-baseline",
+    model_type=["dense"],
     modalities=["text", "image"],
     **_common_mock_metadata,
 )
@@ -233,7 +234,7 @@ class RandomCrossEncoderBaseline:
 random_cross_encoder_baseline = ModelMeta(
     loader=RandomCrossEncoderBaseline,  # type: ignore
     name="baseline/random-cross-encoder-baseline",
+    model_type=["cross-encoder"],
     modalities=["text", "image"],
-    is_cross_encoder=True,
     **_common_mock_metadata,
 )

mteb/models/model_implementations/rasgaard_models.py

@@ -6,6 +6,7 @@ from mteb.models.model_meta import ModelMeta, ScoringFunction
 potion_base_8m = ModelMeta(
     loader=Model2VecModel,  # type: ignore
     name="rasgaard/m2v-dfm-large",
+    model_type=["dense"],
     languages=["dan-Latn"],
     open_weights=True,
     revision="387897cfb09992e6d45ea9cd7b28b9fcf119e23a",

mteb/models/model_implementations/reasonir_model.py

@@ -44,6 +44,7 @@ ReasonIR_8B = ModelMeta(
         trust_remote_code=True,
     ),
     name="ReasonIR/ReasonIR-8B",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="c3d0690370ff4a8c3d3882d8dfa85c43650034fa",

mteb/models/model_implementations/repllama_models.py

@@ -162,6 +162,7 @@ repllama_llama2_original = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="castorini/repllama-v1-7b-lora-passage",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="01c7f73d771dfac7d292323805ebc428287df4f9-6097554dfe6e7d93e92f55010b678bcca1e233a8",  # base-peft revision
@@ -194,6 +195,7 @@ repllama_llama2_reproduced = ModelMeta(
         model_prompts=model_prompts,
     ),
     name="samaya-ai/RepLLaMA-reproduced",
+    model_type=["dense"],
     languages=["eng-Latn"],
     open_weights=True,
     revision="01c7f73d771dfac7d292323805ebc428287df4f9-ad5c1d0938a1e02954bcafb4d811ba2f34052e71",  # base-peft revision