OpenAI sparked the AI revolution with ChatGPT, and DeepSeek emerged because the second family identify within the area. Very quickly, the DeepSeek app topped the iPhone App Retailer’s High Charts, rapidly reaching thousands and thousands of downloads on Google Play Retailer.
Initially, it was all about how good the mannequin was, the way it crashed the market, and the way China had outperformed the US. Now that the hype has cooled down – customers are operating into an issue.
Many customers now report that as an alternative of answering queries, the mannequin regularly responds with: “Server is busy, please strive once more”.
“I relate to DeepSeek at a religious stage as a result of I, too, get an exhausted, busy server after answering two questions,” jested a consumer named Soumya on X.
DeepSeek is ‘Nowhere’ Shut
Whereas this isn’t a problem for customers downloading the mannequin domestically, it’s actually irritating for almost all of individuals utilizing the app.
As an example, Stefano Gogioso, a pc science lecturer on the College of Oxford, mentioned he solely has two judgements relating to DeepSeek. One, it excels within the chain of thought reasoning, and the opposite – the server is busy, strive once more later.
Furthermore, a number of trade professionals are additionally questioning whether or not DeepSeek has what it takes to run the mannequin within the cloud. Dylan Patel, founding father of semiconductor evaluation agency Semi Evaluation, mentioned that DeepSeek is probably going “bleeding out cash”.
“DeepSeek doesn’t have any capability to really serve the mannequin,” he mentioned. “The flexibility to make use of it’s like non-existent now. So many individuals are attempting to make use of it, they don’t have the GPUs to serve it [on cloud],” he added.
DeepSeek reportedly had over 22 million day by day energetic customers on its platform in January, second solely to ChatGPT. Are these too many for a comparatively new AI participant to deal with?
This isn’t simply in regards to the chatbot software; even DeepSeek’s API is dealing with the same destiny, if not worse. Whereas it gives great financial savings over OpenAI, Google, and Anthropic API, it’s experiencing extreme efficiency points. As an example, DeepSeek’s standing exhibits that its API skilled a downtime for 10 steady days.
Is DeepSeek only a Twitter created hype bubble? Their API has been down for 12 hours and nobody has batted an eye fixed.
Ik there’s plenty of suppliers however they’re not as low cost and as quick – issues that individuals had been claiming gave DeepSeek the benefit
Simply fascinating— Pallav Agarwal (@pallavmac) January 27, 2025
Extra just lately, on February 6, it was reported that DeepSeek quickly suspended API top-ups for builders.
Observe that DeepSeek could solely have entry to a portion of the GPUs, and its dad or mum firm, Excessive Flyer, makes use of the remaining for hedge fund actions. That is primarily what the corporate does, and DeepSeek is claimed to be a mere aspect mission.
Earlier, one in all SemiAnalysis’s analysis claims that DeepSeek has entry to 50,000 GPUs. Nevertheless, Patel mentioned that even when one believes their analysis, DeepSeek is ‘nowhere’ near the GPU volumes wanted to serve the mannequin at scale.
In distinction, Patel mentioned that OpenAI and Microsoft have lots of of 1000’s of GPUs between them to serve their high-performance fashions.
An evaluation from Omdia, a tech consulting agency, estimated that in 2024, Microsoft purchased 485,000 items of NVIDIA’s Hopper chips, whereas Meta purchased 224,000 chips. Amazon and Google purchased 196,00 and 169,00 Hopper chips, respectively. That’s the scale at which these corporations function.
These outages is also attributed to the brutal cyber assaults that DeepSeek has confronted. As per the South China Morning Publish, DeepSeek confronted a DDoS (distributed denial of service) assault, with over 230 million requests per second.
Earlier, when DeepSeek-R1 was launched, the Chinese language lab additionally briefly stopped new sign-ups for the platform resulting from a large-scale cyber assault. Thanks to those points, DeepSeek is probably going limiting enter from customers.
In response to a person from the Chinese language state media, these assaults had been reportedly from the US. Finally, what issues to the consumer is the way it mitigates these points. In spite of everything, everybody loves DeepSeek.
However How Do We Repair the ‘Server is busy…’ problem?
Due to its open-source nature, various platforms to make use of DeepSeek can be found in a number of varieties.
The best technique to transfer away from DeepSeek’s native app and nonetheless use the mannequin without cost is thru OpenRouter. This free-to-use platform allows you to entry DeepSeek R1 and the V3 fashions by means of a chat interface.
The opposite various is utilizing DeepSeek on Perplexity. The AI search engine platform permits customers to decide on DeepSeek’s R1 mannequin for responses. The free model of Perplexity gives 5 queries per day with DeepSeek R1.
If you happen to’re sceptical of DeepSeek’s knowledge assortment insurance policies, Perplexity hosts the mannequin on servers in the US.
Lastly, the best choice for utilizing DeepSeek without cost with none server-related points is to obtain the mannequin domestically. A number of sources on the internet present step-by-step directions on how to take action.
For instance, the next information from DataCamp helps you obtain and arrange DeepSeek on Ollama, a platform that allows you to run AI fashions domestically in your system.
The publish Why on Earth is the DeepSeek Server At all times ‘Busy’? appeared first on Analytics India Journal.
