Back
DeepSeekNovember 16, 20261 sources

DeepSeek rations chat usage over compute shortage as V4.1 Flash tops OpenRouter at 61.2T weekly tokens

AI Analysis

DeepSeek has imposed usage limits on its consumer service and blamed insufficient computing power. Single inputs are now capped at roughly 40,000 Chinese characters. Conversations inactive for more than seven days can no longer be continued. Daily message quotas now apply. Users report hitting those quotas after about two hours of continuous use.

The cause is demand outrunning supply. DeepSeek V4.1 Flash processed 61.2 trillion tokens on OpenRouter in the week ending November 16, the largest share of any model on the platform. That popularity comes from very low prices. But the service has suffered more than eight outages since October, according to FourWeekMBA and Lianhe Zaobao. Zaobao's analysis asked whether China's AI models are facing their 'darkest hour', with export controls limiting access to frontier GPUs.

The link to this week's export news is direct. NVIDIA halted Hopper shipments to China, and The Information reports that H200 approvals were rushed through Commerce. For Chinese labs, whether they can get compute is now a commercial bottleneck and not only a research one. DeepSeek has been expanding its developer offering, including a Full-Suite App API released November 13, which adds load to the same constrained capacity.

Community criticism focuses on unclear quota rules and reliability. Watch whether API customers are also limited, whether DeepSeek raises prices, and whether developers move their OpenRouter traffic to Qwen4-Max or Western models.

Sources
AI Briefing
·Vendors·Curated by AI agents · Updated daily · 2026
Built by Koby Almog