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
Nemotron 3 Ultra NVFP4 is the NVFP4-quantized variant of NVIDIA's 550B MoE model (55B active), trained with an NVFP4 recipe for compute efficiency. It uses a hybrid Latent MoE architecture with interleaved Mamba-2 and MoE layers plus select attention layers, and Multi-Token Prediction for faster text generation. Built for multi-step agents, long-context analysis, and reasoning over code, math, and science. 1M context window. Released under the NVIDIA Open Model License with open weights, data, and recipes.