#7 Nemotron 3.5 Super VL
The most document-focused model here, first-party DGX Spark (NVFP4) support, and a matching embed/rerank stack.
- Version
- NVIDIA-Nemotron-3.5-Super-VL-120B-A12B
- Maker
- NVIDIA
- Release date
- no published figure yet unverified (Hugging Face repo creation date unless noted)
- Architecture
- Hybrid Mamba-2 + MoE + attention, C-RADIO vision encoder
- Total params
- 120B
- Active params
- 12B
- Context length
- Up to 1M
- License
- OpenMDW-1.1
- Commercial notes
- Open license; read terms
- Variants
- BF16
- NVFP4 (DGX Spark recommended)
- Base-BF16
- Nemotron 3.5 Lightning 30B-A3B
- Bolt 74B-A9B
- Memory
- BF16/native: no published figure yet
FP8: no published figure yet
Q8: no published figure yet
Q4: no published figure yet
BF16 deploy: 4x H100 or 2x GB200; NVFP4 card targets DGX Spark - Hardware
- Single Spark (128GB): Yes via NVFP4 (NVIDIA card points Spark users to it)
Dual Spark (256GB): Yes
Quad Spark (~512GB): Yes
24GB GPU: No
128GB Mac / Strix: no published figure yet - Tokens/sec
- api_artificial_analysis: 449 tok/s (hosted, not local)
- Document Q&A strengths
- Built for document understanding: MMLongBench-Doc 62.82 (model card)
- Docs as page images + text
- 1M context
- Embedding / reranker pairing
- {'embedding': 'nvidia/Nemotron-3.5-Embed-8B or -1B', 'reranker': 'nvidia/Nemotron-3.5-Rerank-8B (SOTA ViDoRe v3.1 reranking, 2026-10-20)', 'source': 'https://huggingface.co/nvidia/Nemotron-3.5-Rerank-8B-BF16', 'verified': 'same-family pairing; NVIDIA does not explicitly prescribe this combo'}
- Runtimes
- Ollama: listed (nemotron-3.5-super) unverified
llama.cpp: no published figure yet
vLLM: Yes
LM Studio: no published figure yet - Hugging Face
- https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16
DGX Spark configs it fits
- Single Spark (128GB): Yes via NVFP4 (NVIDIA card points Spark users to it)
- Dual Spark (256GB): Yes
- Quad Spark (~512GB): Yes
Spark fit is estimated from published memory sizes, not tested on Spark hardware.
Sources
- https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16
- https://huggingface.co/nvidia/Nemotron-3.5-Rerank-8B-BF16
- https://theopenweights.com/leaderboards/llm-intelligence
Last updated Oct 24, 2026. Fields tagged unverified lack a primary source. "No published figure yet" means we found no source, and we don't guess. Spark fit is estimated from published memory sizes, not tested on Spark hardware. Ollama listings may be cloud-only tags; check this for Kimi K3 and DeepSeek V4.1 before pulling.
The RAGShelf weekly
New local RAG tools, releases, and tested setups. One email a week, no spam.