Modelofficial release

nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16

NVIDIA Nemotron 3 →Hugging Face ↗
Parameters124B (12B active)
Context1M
Input → outputimage + text → text
Licenseother
PrecisionBF16
ArchitectureNemotronH_Omni_Reasoning_V3
ReleasedOct 12, 2026
Downloads142

Other sizes and formats in the family

ModelParamsPrecisionDownloads
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-Base-BF16124BBF164
nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF1674.9BBF164
nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-GGUFgguf74B–119
nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-LiLiCorr-GGUFgguf74B–42
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4nvfp465.8BU881
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Base-BF1631.6BBF1631.8K
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF1631.6BBF16692.3K
nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GBquantized22.4BU858
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4nvfp417.8BU8937.7K
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4-LiLiCorrnvfp4965MU8137
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4-DSparknvfp4764MBF16150K
nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-LiLiCorrquantized673MU820
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4-DFlashnvfp4663MBF161.4K

Compare side by side →

Built on this model

ModelParamsPrecisionDownloads
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4-LiLiCorrnvfp4965MU8137
nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF1674.9BBF164

This release in the news

  1. Oct 15, 2026 · Model release
    nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4-LiLiCorr

Recent news about NVIDIA Nemotron 3

  1. Oct 19, 2026 · Opinion / analysis
    Kapitsa: open-source Apple Silicon optimized Metal engine built for Qwen and Nemotron. See benchmarks of Qwen 27B, Nemotron + 8 other models on an M1 Pro (16GB), M4 Pro (48GB) and M5 Max (128GB)
  2. Oct 16, 2026 · Model release
    Tyger: open-weights streaming ASR + full-duplex turn-control in one model
  3. Oct 15, 2026 · Opinion / analysis
    nvidia/Nemotron-3.5-Embed-1B-BF16 · Hugging Face
  4. Oct 15, 2026 · Model release
    NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal Reasoning
  5. Oct 15, 2026 · Open-source release
    unslothai/unsloth v0.1.908-beta: NVIDIA Nemotron 3.5 + Meta Muse Spark Open
  6. Oct 15, 2026 · Opinion / analysis
    Add Nemotron Super 3.5 VL and Nemotron Bolt 74B A9B support by frozenblade1224 · Pull Request #30520 · ggml-org/llama.cpp
  7. Oct 15, 2026 · Model release
    nvidia/Nemotron-3.5-Rerank-8B-BF16
  8. Oct 12, 2026 · Research paper
    ReplaySSM: Rethinking State Storage for Faster Hybrid-Model Decoding