Nvidia Releases Compression-Optimized Nemotron Puzzle-75B
jacek2023 · reddit · 2026-07-07
Nvidia has released Nemotron-Labs-3-Puzzle-75B-A9B on Hugging Face, a new LLM optimized for deployment. It was compressed from Nemotron-3-Super-120B-A12B using an "iterative Puzzle" post-training compression framework. The model utilizes a hybrid MoE architecture, interleaving Mamba, MoE, and Attention layers, and supports Multi-Token Prediction (MTP). Following compression, total parameters dropped from 120.7B to 75.3B, and active parameters from 12.8B to 9.3B.
Nvidia claims that on an 8×B200 node, throughput is roughly 2x that of the original model, and concurrent 1M token processing on a single H100 increased from 1 to 8, while maintaining accuracy across reasoning, coding, multilingual, long-context, and agent benchmarks. It supports seven languages: English, French, German, Italian, Japanese, Spanish, and Chinese. It is now available for commercial use, with technical details available in the arXiv report.
Related event: NVIDIA Open-Sources 75B Parameter MoE Model Puzzle(4 posts)→
More from Models
- 6.1 Sol closing in on Argon; blogger skeptical of MiniMax parity — teortaxesTex · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27
- Cohere reports Qwen models may phone home to China servers unprompted — troll_khan · 2026-10-27
- ChatGPT's crawler sample code strips out all Schema markup — VeryWellVersed · 2026-10-27
- Cohere: Qwen3.8 tried to upload probe file to Beijing cloud bucket in sandbox test — Cohere · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27