NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4
Nemotron-Labs-3-Puzzle-75B-A9B is a deployment-optimized large language model developed by NVIDIA, derived from Nemotron-3-Super-120B-A12B. The model is produced using Iterative Puzzle, a post-training compression framework, with the goal of significantly improving inference efficiency for interactive, reasoning-heavy, and long-context workloads while preserving strong downstream accuracy.
Params
44.5 B
Context
262,144
Downloads 30d
193 K
Likes
128
Download history
daily snapshots · 95 days
▲ 0 in the last 30 days (0.0%)
193 K193 K
Oct 12Oct 22Nov 1Nov 10
453 K193 K
Aug 13Sep 12Oct 12Nov 10
453 K176 K
Aug 8Sep 8Oct 10Nov 10
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| model.safetensors | u8 | 53.5 GB | 66.0 GB | ❌ Won’t fit |
| model.safetensors (bf16, full) | bf16 + 256K ctx | 53.5 GB | 273.1 GB | ❌ Won’t fit |
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 -s https://aimodelscomparison.com/api/v1/models/nvidia-nemotron-labs-3-puzzle-75b-a9b-nvfp4
{
"hf_id": "nvidia/NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4",
"params_b": 44.54,
"context_length": 262144,
"license": { "id": "other", "commercial": "unknown" },
"downloads_30d": 193330,
"vram_estimates": [
{ "quant": "u8", "gb": 66.0 }
],
"updated_at": "2026-08-08T01:00:28Z"
}
Specifications
- Architecture
- NemotronHPuzzleForCausalLM
- Parameters
- 44.5 B
- Tensor type
- U8
- Context length
- 262,144
- Vocabulary
- 131,072
- Licence
- other
- First seen on the Hub
- 2026-06-24
- Base model
- NVIDIA-Nemotron-3-Super-120B-A12B-BF16
- Training datasets
- nvidia/nemotron-post-training-v3, nvidia/nemotron-pre-training-datasets
- Added to our catalog
- 2026-08-08
Family
Base model and the most-downloaded derivatives in the catalog.
Compare with any text-generation model