nvidia / text-generation updated 2 weeks ago

NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4-DSpark

The NVIDIA Nemotron-3.5-Lightning-30B-A3B-NVFP4-DSpark model is the DSpark speculative decoding checkpoint for NVIDIA's Nemotron-3.5-Lightning-30B-A3B model family, which is a hybrid LatentMoE language model designed for reasoning, chat, and agentic workflows. For more information, please check BF16, NVFP4. The NVIDIA Nemotron-3.5-Lightning-30B-A3B-DSpark-NVFP4 model is intended for DSpark speculative decoding deploy...

Params
760 M
Context
1,048,576
Downloads 30d
241 K
Likes
30
Licence: other Not gated SAFETENSORS View on Hugging Face ↗

Download history

daily snapshots · 22 days
326 K191 K
Aug 31Sep 7Sep 14Sep 21

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

FileQuantSizeEst. VRAMVerdict on RTX 4090 · 24 GB
model.safetensors bf16 1.3 GB 2.1 GB ✅ Runs comfortably
model.safetensors (bf16, full) bf16 + 1024K ctx 1.3 GB 16.6 GB ✅ Runs comfortably
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Run it

copy-paste, exact tags checked against the Hub
~ · curl · api/v1
$ curl -s https://aimodelscomparison.com/api/v1/models/nvidia-nemotron-3-5-lightning-30b-a3b-nvfp4-dspark
{
  "hf_id": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4-DSpark",
  "params_b": 0.76,
  "context_length": 1048576,
  "license": { "id": "other", "commercial": "unknown" },
  "downloads_30d": 241067,
  "vram_estimates": [
    { "quant": "bf16", "gb": 2.1 }
  ],
  "updated_at": "2026-09-02T01:00:29Z"
}
est. VRAM —on RTX 4090 · 24 GBJSON API →

Specifications

Architecture
Qwen3DSparkModel
Parameters
760 M
Tensor type
BF16
Context length
1,048,576
Vocabulary
131,072
Layers / heads
6 / 32
Licence
other
First seen on the Hub
2026-08-05
Base model
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16
Training datasets
undisclosed
Added to our catalog
2026-08-31

Family

Base model and the most-downloaded derivatives in the catalog.

Compare with any text-generation model