GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
nvidia
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
nvidia
Model Tree
Input Modalities
Output Modalities
Supported Functionality
NVIDIA Nemotron 3 Ultra is a 550B MoE model (55B active) for long-running autonomous agents in coding, deep research, and enterprise workflows. Its hybrid Mamba-Transformer architecture pairs Latent MoE (4 experts at the cost of 1) with Multi-Token Prediction and Token Budget control for cost-managed reasoning. A 1M context window supports sustained agent sessions. It generates a reasoning trace before the final answer, configurable via chat template. NVIDIA Open Model License with open weights, data, and recipes.