Support matrix v7

Local embedded models

The following HuggingFace model variants have been tested and verified to work with AIDB's local model providers. You can use any of these identifiers with the corresponding config helper when calling aidb.create_model().

BERT

Model nameHuggingFace identifierConfig string
bert_all-MiniLM-L12-v1sentence-transformers/all-MiniLM-L12-v1aidb.bert_config('sentence-transformers/all-MiniLM-L12-v1', revision => 'main')
bert_all-MiniLM-L6-v1sentence-transformers/all-MiniLM-L6-v1aidb.bert_config('sentence-transformers/all-MiniLM-L6-v1', revision => 'main')
bert_all-MiniLM-L6-v2sentence-transformers/all-MiniLM-L6-v2aidb.bert_config('sentence-transformers/all-MiniLM-L6-v2', revision => 'main')
bert_all-distilroberta-v1sentence-transformers/all-distilroberta-v1aidb.bert_config('sentence-transformers/all-distilroberta-v1', revision => 'main')
bert_msmarco-bert-base-dot-v5sentence-transformers/msmarco-bert-base-dot-v5aidb.bert_config('sentence-transformers/msmarco-bert-base-dot-v5', revision => 'main')
bert_multi-qa-MiniLM-L6-cos-v1sentence-transformers/multi-qa-MiniLM-L6-cos-v1aidb.bert_config('sentence-transformers/multi-qa-MiniLM-L6-cos-v1', revision => 'main')
bert_multi-qa-MiniLM-L6-dot-v1sentence-transformers/multi-qa-MiniLM-L6-dot-v1aidb.bert_config('sentence-transformers/multi-qa-MiniLM-L6-dot-v1', revision => 'main')
bert_paraphrase-TinyBERT-L6-v2sentence-transformers/paraphrase-TinyBERT-L6-v2aidb.bert_config('sentence-transformers/paraphrase-TinyBERT-L6-v2', revision => 'main')
bert_paraphrase-multilingual-MiniLM-L12-v2sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2aidb.bert_config('sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2', revision => 'main')
bert_paraphrase-multilingual-mpnet-base-v2sentence-transformers/paraphrase-multilingual-mpnet-base-v2aidb.bert_config('sentence-transformers/paraphrase-multilingual-mpnet-base-v2', revision => 'main')

All BERT variants support text embedding (aidb.encode_text()).

CLIP

Model nameHuggingFace identifierConfig string
openai/clip-vit-base-patch32openai/clip-vit-base-patch32aidb.clip_config('openai/clip-vit-base-patch32')

CLIP supports both text embedding (aidb.encode_text()) and image embedding (aidb.encode_image()).

Llama

Model nameHuggingFace identifierConfig string
HuggingFaceTB/SmolLM2-1.7B-InstructHuggingFaceTB/SmolLM2-1.7B-Instructaidb.llama_config('HuggingFaceTB/SmolLM2-1.7B-Instruct', revision => 'main')
HuggingFaceTB/SmolLM2-135M-InstructHuggingFaceTB/SmolLM2-135M-Instructaidb.llama_config('HuggingFaceTB/SmolLM2-135M-Instruct', revision => 'main')
HuggingFaceTB/SmolLM2-360M-InstructHuggingFaceTB/SmolLM2-360M-Instructaidb.llama_config('HuggingFaceTB/SmolLM2-360M-Instruct', revision => 'main')
TinyLlama/TinyLlama-1.1B-Chat-v1.0TinyLlama/TinyLlama-1.1B-Chat-v1.0aidb.llama_config('TinyLlama/TinyLlama-1.1B-Chat-v1.0', revision => 'main')

All Llama variants support text generation (aidb.decode_text()).

T5

Model nameHuggingFace identifierConfig string
t5-smallt5-smallaidb.t5_config('t5-small', revision => 'main')
t5-baset5-baseaidb.t5_config('t5-base', revision => 'main')
t5-larget5-largeaidb.t5_config('t5-large', revision => 'main')

T5 variants support both text encoding (aidb.encode_text()) and text generation (aidb.decode_text()).

Note

These are the pre-vetted variants confirmed to work with AIDB. You are not limited to these — any compatible HuggingFace model can be used with the appropriate config helper, but only the variants listed here have been tested by EDB.


NIM models

AIDB supports NVIDIA NIM microservices via several providers. The tables below list the compatible NIM models for each provider. Models marked as Default on HM are shipped with EDB Hybrid Manager's Model Serving and are available out of the box in HM deployments — no separate deployment required.

NIM completions (nim_completions)

Supports: aidb.decode_text(), aidb.decode_text_batch()

NIM model identifierDefault on HM
nvidia/nemotron-3-nano
nvidia/llama-3.3-nemotron-super-49b-v1
ibm/granite-guardian-3.0-8b
ibm/granite-3.0-8b-instruct
ibm/granite-3.0-3b-a800m-instruct
meta/llama-3.3-70b-instruct
meta/llama-3.2-3b-instruct
meta/llama-3.2-1b-instruct
meta/llama-3.1-405b-instruct
meta/llama-3.1-70b-instruct
meta/llama-3.1-8b-instruct
meta/llama3-70b-instruct
meta/llama3-8b-instruct
nvidia/llama-3.1-nemotron-70b-instruct
nvidia/llama-3.1-nemotron-51b-instruct
nvidia/nemotron-mini-4b-instruct
nvidia/nemotron-4-340b-instruct
google/shieldgemma-9b
google/gemma-7b
google/codegemma-7b

NIM embeddings (nim_embeddings)

Supports: aidb.encode_text(), aidb.encode_text_batch()

NIM model identifierDefault on HM
nvidia/llama-3.2-nemoretriever-300m-embed-v1
nvidia/nv-embedqa-e5-v5

NIM CLIP (nim_clip)

Supports: aidb.encode_text(), aidb.encode_text_batch(), aidb.encode_image()

NIM model identifierDefault on HM
nvidia/nvclip

NIM OCR (nim_paddle_ocr)

Supports: aidb.perform_ocr()

NIM model identifierDefault on HM
baidu/paddleocr

NIM reranking (nim_reranking)

Supports: aidb.rerank_text()

NIM model identifierDefault on HM
nvidia/llama-3.2-nv-rerankqa-1b-v2
Note

Models marked as Default on HM are pre-deployed by EDB Hybrid Manager's Model Serving. All other models can be self-hosted using KServe and registered with AIDB using aidb.create_model(). For setup instructions, see External model connections.