Source
NVIDIA Vera Rubin Maximizes Intelligence per Dollar for Post-Training | NVIDIA Blog
Lowest cost per token from extreme codesign maximizes intelligence per dollar for post-training in the agentic era.
Show entities and relationshipsHide entities and relationships
In this article
Products
Technologies
Key connections
Prime Intellect uses NVIDIA Blackwell
Prime Intellect continuously post-trains open models on NVIDIA Blackwell.
NVIDIA owns NVIDIA Dynamo
NVIDIA provides Dynamo system software for AI factory disaggregation.
Prime Intellect uses NVIDIA Dynamo
Prime Intellect uses NVIDIA Dynamo for inference orchestration.
Show 12 more connectionsShow fewer connections
Prime Intellect uses NVIDIA Vera CPU
Prime Intellect optimizes its sandbox infrastructure for NVIDIA Vera CPUs.
Perplexity uses NVIDIA GB200 NVL72
Perplexity serves post-trained models on NVIDIA GB200 NVL72 systems.
Together AI uses NVIDIA Vera Rubin
Together AI aims to harness the NVIDIA Vera Rubin platform for post-training.
NeMo Gym is related to Reinforcement Learning
NeMo Gym provides training environments for reinforcement learning.
NeMo RL is related to Reinforcement Learning
NeMo RL enables distributed reinforcement learning for post-training.
NVIDIA Nemotron 3 Ultra is built with Mixture of Experts
NVIDIA Nemotron 3 Ultra is a 550-billion-parameter mixture-of-experts model.
NVIDIA Nemotron 3 Ultra is related to SWE-bench Verified
NVIDIA Nemotron 3 Ultra scored 71.7% on the SWE-bench verified benchmark.
Perplexity uses RDMA
Perplexity uses an RDMA-based weight transfer engine.
Together AI is related to Supervised Fine-Tuning
Together AI offers supervised fine-tuning as part of its post-training service.
Together AI is related to Direct Preference Optimization
Together AI supports direct preference optimization in its post-training suite.
Agentic AI is related to Reinforcement Learning
Agentic AI models rely on reinforcement learning techniques to adapt and learn.
NVIDIA owns Nemotron 3 Super
NVIDIA developed and released the Nemotron 3 Super model.
Get the weekly recap
The stories like this one, picked and explained — once a week, straight to your inbox.