BREAKING
NVIDIA Releases Nemotron 3
0B
Nano total
0B
Super total
0B
Ultra total
Hybrid design, open recipes
ArchitectureHybrid
Mamba-2 for long context
Attention for accuracy
1M-token context
OpennessOpen
Weights published
Training data shared
Commercial use allowed
Nano throughput vs rivals
vs Qwen3-30B3.3
vs GPT-OSS-20B2.2
Ultra throughput edge
vs GLM-5.15.9
vs Kimi-K2.64.8
vs Qwen-3.51.6
Open weights on Hugging Face
AI NEWS BLITZ
NVIDIA has launched Nemotron 3, a family of open-weight large language models.