DeepSeek's New LLM Engine Just Made GPUs 3x More Valuable
bycloud · 19:18 · Watch on YouTube
DeepSeek’s V4.1 Flash 1017 update and the open-sourced Deep Infer engine are presented as the reason token revenue per GPU jumped from about $12.91/hour on the February 2025 V3/R1 stack to about $38/hour off-peak: pre...