#Blackwell

NVIDIA AI published Day 0 performance benchmarks for DeepSeek-V4-Pro (1M long-context model) running on Blackwell Ultra infrastructure, using vLLM's Day 0 recipe. The release establishes baseline throughput vs interactiv…

NVIDIA announces DeepSeek-V4, a 1.6T parameter LLM with million-token context window optimized for agentic workflows. Running on Blackwell Ultra hardware, the model achieves over 150 TPS/user throughput. Future performan…