DeepSeek Limits Chat Length Amidst Compute Strain

0
10

Users of the AI platform DeepSeek have recently reported that the service has begun imposing limits on its long chat functionality across both its application and web interfaces. This move comes as the company grapples with significant strain on its computing resources, a situation that has been escalating in recent months.

Usage Restrictions Introduced

Multiple users have received notifications from the platform indicating that long conversations inactive for more than seven days will no longer be supported, citing “insufficient computing power.” Additionally, the maximum length for a single text input has been reduced to approximately 40,000 characters. A daily usage cap has also been put in place, with users encountering a “Usage limit reached due to server congestion” message once their daily allowance is depleted.

As of the time of reporting, DeepSeek has not issued a formal statement addressing these new limitations or provided an estimated timeline for their removal.

Ongoing Compute Challenges

These new restrictions appear to be a direct consequence of DeepSeek’s ongoing struggles with compute capacity. The company’s official status page reveals a history of service disruptions. Since October 2026, there have been eight reported outages affecting both its conversational services and its API offerings, specifically V4 Pro and V4.1 Flash. During the period from August to October, the V4.1 Flash API experienced an availability of 99.46%, translating to approximately 47.3 hours of downtime over the entire year.

Surging Demand Exacerbating Strain

The significant increase in demand for DeepSeek’s services is a primary driver behind the current pressure on its infrastructure. Data from OpenRouter, a platform aggregating multiple AI models, highlights a dramatic surge in usage for DeepSeek’s V4.1 Flash model. During the week of October 19th to 25th, the model processed an astonishing 33.6 trillion tokens, marking an 801% increase week-over-week. This represents a more than fourfold increase in just under three months, compared to the 7.22 trillion tokens processed in late July and early August.

This rapid growth in user engagement and computational demand has clearly outpaced DeepSeek’s current infrastructure capacity, leading to the implementation of these usage controls. The situation underscores the challenges faced by AI companies in scaling their operations to meet the rapidly growing global demand for advanced AI models and services.

This development follows earlier observations by developers who noticed “quota exhausted” messages in DeepSeek’s front-end code, leading to speculation about the potential introduction of a paid tier for end-users.

Source: https://www.ithome.com/1/027/597.htm

LEAVE A REPLY

Please enter your comment!
Please enter your name here