Overview
From Hybrid Manager, there are two primary consumers of AI models:
- PG.AI Knowledge Base (AIDB Postgres extension) for creating and maintaining AI Knowledge Bases.
- Langflow for building AI pipelines and agentic flows.
Default NIM Models
| Model type | NIM model | NVIDIA NIM documented resource requirements |
|---|---|---|
| Text completion | nvidia/llama-3.3-nemotron-super-49b-v1 | 4 × L40S |
| Text embeddings | nvidia/llama-3.2-nemoretriever-300m-embed-v1 | 1 × L40S |
| Image embeddings | nvclip | 1 × L40S |
| OCR | paddleocr | 1 × L40S |
| Text reranking | llama-3.2-nv-rerankqa-1b-v2 | 1 × L40S |
Minimum GPU Requirement
Based on the default models above, the minimum to run them concurrently is 8 × L40S GPUs.
Cloud Mappings
- AWS EKS: recommend a node group with 2 ×
g6e.12xlargenodes. - GCP GKE: recommend a node pool with 2 ×
a2-highgpu-4gnodes.
Note: GCP does not offer L40S GPUs. The recommended A2 nodes with A100 GPUs are supported and documented for the NIM models listed above.
Additional Supported NIM Models
Beyond the defaults above, AIDB and Langflow support the following additional NIM model variants, tested and confirmed to work with Agent Factory. Models marked Default on HM are the same ones already covered by the GPU sizing above; all others can be self-hosted with KServe and registered the same way, but aren't included in the minimum GPU requirement or cloud mappings on this page — consult the model's page on build.nvidia.com for its own resource requirements before deploying it.
Text Completion (nim_completions)
| NIM model | Default on HM | Description |
|---|---|---|
meta/llama-3.1-8b-instruct | — | General-purpose instruction-following chat model; solid default choice for most completion tasks. |
nvidia/llama-3.3-nemotron-super-49b-v1 | ✓ | Large reasoning/chat model for complex tasks needing higher-quality output. |
nvidia/llama-3.3-nemotron-super-49b-v1.5 | — | Refined revision of llama-3.3-nemotron-super-49b-v1 with improved response quality. |
nvidia/nemotron-3-nano | ✓ | Compact, fast completion model — good for latency-sensitive workloads. |
nvidia/nvidia-nemotron-nano-9b-v2 | — | Mid-size reasoning-capable model, balancing speed and capability. |
openai/gpt-oss-120b | — | Large open-weight model for high-quality completions and agentic tasks. |
openai/gpt-oss-20b | — | Smaller open-weight model — faster, lighter-weight completions. |
Text Embeddings (nim_embeddings)
| NIM model | Default on HM | Description |
|---|---|---|
nvidia/llama-3.2-nemoretriever-300m-embed-v1 | ✓ | General-purpose text embeddings for retrieval and RAG. |
nvidia/llama-3.2-nv-embedqa-1b-v2 | — | Larger embedding model tuned for question-answering retrieval. |
nvidia/nv-embedqa-e5-v5 | — | Embedding model tuned specifically for QA-style retrieval tasks. |
Image Embeddings (nim_clip)
| NIM model | Default on HM | Description |
|---|---|---|
nvidia/nvclip | ✓ | Joint text/image embeddings for multimodal search. |
OCR (nim_paddle_ocr)
| NIM model | Default on HM | Description |
|---|---|---|
baidu/paddleocr | ✓ | General-purpose OCR text extraction from images. |
nvidia/nemoretriever-ocr-v1 | — | NVIDIA's OCR model — an alternative to PaddleOCR. |
Text Reranking (nim_reranking)
| NIM model | Default on HM | Description |
|---|---|---|
nvidia/llama-3.2-nv-rerankqa-1b-v2 | ✓ | Reranks retrieved passages by relevance to a query — use after an initial vector search to improve top-k precision. |