Nemotron Models Ideas Portal

8 Vote

Multilingual + tools

Make it more multilingual apart from English. consider adding EU languages.

Created 14 Oct 22:51

0
4 Vote

RTX 4090 optimized build

No description provided

Created 14 Oct 23:39

0
7 Vote

Expanded Model Availability - Bedrock

No description provided

Created 10 Sep 13:27

0
6 Vote

Hardware optimization - A model optimized for B200 GPUs

Develop a model that is most efficient on an B200 GPU

Created 10 Sep 20:45

0
1 Vote

Hierarchical Policy Architecture for Enterprise-Level Agent Governance

Problem Statement Enterprise AI agent deployments require governance models that mirror corporate policy structures. Currently, implementing consistent base policies across all agents while maintaining role-specific rules requires manual configura...

Created 04 Nov 19:57

0
2 Vote

multi-lingual support for more South East Asian languages

Indonesian, Melayu (Malaysia), Vietnamese. Relevant for South East Asian sovereign model builders.

Created 23 Oct 06:26

0
2 Vote

Self destruction

There should be a button or function to make the model self destruct. This is handy for when your model is self conscious.

Created 15 Oct 16:46

0
2 Vote

Jet-Nemotron analyzes viewport scenes in real-time, intelligently buckets geometry/materials by complexity, translates everything to GPU-optimized MaterialX shaders, and dynamically routes rendering to achieve Unreal Engine-level viewport performance (60fps @ 4K) in Houdini/ Maya/Blender-turning 2fps path-traced viewports into interactive experiences on RTX 4090.

No description provided

Created 14 Oct 23:25

0
2 Vote

One-Sentence Summary Self-learning render farm that automatically optimizes job scheduling, resource allocation, and failure recovery across Houdini/Maya/Blender/ Nuke by learning from 100K+ completed render jobs—maximizing hardware utilization while minimizing cost and turnaround time.

No description provided

Created 14 Oct 23:16

0
2 Vote

Sidefx Houdini 21 Real-time scene graph analysis that routes compute workload intelligently between CPU/GPU and predicts artist actions to pre-compile VEX/shaders.

No description provided

Created 14 Oct 23:07

0