DeepSeek Limits Free API Usage, Directs Heavy Users to Paid Plans

iconTechFlow
Share
AI summary iconSummary
AI + crypto news: DeepSeek has restricted free API access, limiting inactive long conversations and capping daily input at 40,000 characters. Users now encounter hidden quotas and “server busy” warnings. The company has also launched an API platform for developers, featuring tools such as web search and file reading. An open-source demo client is available with personal API keys. On-chain developments suggest a move toward paid plans, though no official pricing or free usage limits have been confirmed for the rumored 'Coding Plan' subscription.

Author: Shenchao TechFlow

On November 16, AI unicorn DeepSeek implemented multi-dimensional restrictions on the free services of its official dialogue platform. Several users reported that their mobile app and web interface no longer support long conversations with no interaction for more than 7 days, and the character limit per input has been reduced to approximately 40,000 characters.

Meanwhile, the platform has begun implementing implicit quotas on daily conversations, triggering an interception message reading “Usage limit reached due to high server load” once the threshold is exceeded. As massive numbers of end users continue to flood in, this technology unicorn known for its exceptional value is facing severe challenges in balancing inference compute consumption with commercialization.

Mining costs are rising, and there is speculation that frontend code has been leaked, suggesting paid subscriptions.

The sudden tightening of free services was not without warning. Previously, technical developers had discovered a complete quota management system within DeepSeek’s frontend code, which automatically synchronizes remaining API quotas for each model every five minutes and displays clear alerts to frontend users when core computing resources are exhausted.

More notably, the frontend code explicitly includes a field named "Coding Plan" for subscription packages. In an industry where intensive inference computing costs are high, a no-strings-attached, fully free computing subsidy model is unsustainable—this code detail has sparked widespread market speculation about an upcoming personal paid subscription service.

From industry practice, major large model providers typically implement tiered rate limiting and subscription memberships after experiencing explosive user growth, to balance server bandwidth and computational costs. DeepSeek’s move indicates its attempt to gradually curb uncontrolled computational consumption while ensuring basic availability, paving the way for future commercialization efforts.

Open application-level APIs to build a developer ecosystem adoption pathway

While reducing free traffic, DeepSeek has long established on-ramps for developers. On November 13, the DeepSeek Open Platform officially opened its full suite of APIs to technical developers, including web search, file reading, web content scraping, and multimodal voice capabilities.

In conjunction with this action, the official has simultaneously open-sourced a lightweight demo client, chat-demo, on GitHub. High-frequency users affected by free tier limits can export their conversation history in full and continue using the service by deploying this open-source client locally and connecting it to their personal API key.

According to the billing rules in DeepSeek's official technical documentation, switching to the API call model will result in strict pay-as-you-go pricing based on actual Token consumption. This product strategy of tightening one side while loosening the other alleviates high-concurrency pressure on the official free servers while channeling heavy productivity users into the developer billing system.

As of now, DeepSeek has not publicly disclosed the daily free conversation quota limit, nor has it officially announced the pricing or launch date for the "Coding Plan" subscription package. Amid the tension between commercializing large models and providing accessible services, how to smoothly navigate the period of computational strain remains a shared challenge for the entire industry.

Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.