DeepSeek tightens free quota under pressure on computing power, guiding heavy users to switch to API.

CN
13 minutes ago
The large model unicorn tightens free quotas for C-end users, simultaneously open-sourcing a lightweight client to guide high-frequency users towards a usage-based billing developer ecosystem.

Author: Deep Tide TechFlow

On November 16, the AI unicorn DeepSeek officially implemented multi-dimensional restrictions on its official chat platform free services. Multiple users reported in their daily interactions that its mobile app and web version no longer support long conversations without interaction for over 7 days, and the character limit for a single input has also been compressed to about 40,000 characters.

Meanwhile, the platform has started to set implicit quotas for daily conversations, and a notification stating "Due to server congestion, the usage quota has reached the limit" will pop up after triggering the calling threshold. With a continuous influx of massive C-end users, this technology unicorn, known for its extreme cost-performance ratio, is facing severe challenges in balancing inference computing power consumption and commercialization.

Rising computing costs, speculation on paid subscription plans from front-end code leakage

The sudden tightening of free services did not come without warning. Previously, some tech developers discovered a complete set of quota management mechanisms in the DeepSeek web front-end code. This mechanism can automatically sync the remaining calling quotas for each model every 5 minutes and output clear reminders to front-end users when core computing resources are exhausted.

More notably, the front-end engineering code explicitly mentions a subscription package field named "Coding Plan." In the context of high costs for intensive inference computing power, a barrier-free, completely free computing subsidy model is difficult to sustain, and this coding detail has sparked widespread discussion in the market about its upcoming personal paid subscription service.

From industry practice, mainstream large model vendors typically use tiered throttling and membership subscriptions to balance server bandwidth and computing expenses after experiencing explosive user growth. DeepSeek's actions indicate that it is attempting to gradually consolidate disorderly computing consumption while ensuring basic usability, paving the way for the next steps in commercialization.

Open application-level APIs, building a developer ecosystem pathway

While shrinking free traffic, DeepSeek has already laid a pathway for developers. On November 13, the DeepSeek open platform officially opened up a complete suite of APIs for internet search, file reading, web content scraping, and multimodal voice to technical developers.

Alongside this move, the official GitHub community has simultaneously open-sourced a lightweight demo client chat-demo. High-frequency users affected by the free quota limits can fully export their historical conversation records and then deploy this open-source client locally to access their personal interface token to continue using the service.

According to the billing rules in the official DeepSeek technical documentation, after switching to the interface calling model, payment will be strictly based on the actual token consumption. This product strategy of charging and refunding eases the high concurrency pressure on the official free servers while retaining heavy productivity users within the developer billing system.

Currently, DeepSeek has not disclosed the specific quota limit for daily free conversations, nor has it officially announced the specific pricing and launch timeline for the "Coding Plan" subscription package. In the tug-of-war between large model commercialization exploration and inclusive services, how to smoothly navigate the period of computing pressure remains a common challenge facing the entire industry.

免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到[email protected],本平台相关工作人员将会进行核查。

Share To
APP

X

Telegram

Facebook

Reddit

CopyLink