BREAKING
NVIDIA Releases Nemotron 3
0
B
Nano total
0
B
Super total
0
B
Ultra total
Hybrid design, open recipes
Architecture
Hybrid
●
Mamba-2 for long context
●
Attention for accuracy
●
1M-token context
Openness
Open
●
Weights published
●
Training data shared
●
Commercial use allowed
Nano throughput vs rivals
vs Qwen3-30B
3.3
vs GPT-OSS-20B
2.2
Ultra throughput edge
vs GLM-5.1
5.9
vs Kimi-K2.6
4.8
vs Qwen-3.5
1.6
Open weights on Hugging Face
AI NEWS BLITZ
NVIDIA has launched Nemotron 3, a family of open-weight large language models.