DeepSeek has taken the AI industry (and the stock market) by storm, with its highly competitive V3 and R1 models, that have already been downloaded more than 700,000 times by developers and organizations alike. It is not just fast, but super cost-efficient, suddenly making developers rethink their allegiance to the LLMs that have been around for a while. The question remains, should we get swayed by low prices just yet? Read on to learn more about this tech disruptor and how you can choose the right LLM for your project!

Introduction to DeepSeek

Overlapping the announcement of the Stargate Project, DeepSeek released its R1 LLM, an open-sourced, reasoning model that’s competing head-to-head with OpenAI’s top models at 25% of the cost. Signaling a major shift in AI accessibility and cost efficiency, DeepSeek’s models promise affordability and efficiency.

Within just a few days of its release, an AI assistant app from the house of DeepSeek has already dethroned OpenAI’s ChatGPT from the Apple App Store, getting a statement from Sam Altman, Founder of OpenAI, saying it did a “couple of nice things” but has been “wildly overstated.”

Why DeepSeek

The secret lies in the following factors;

Compared to models from OpenAI, Anthropic, and other industry leaders currently, DeepSeek’s pricing model is so much more attractive, making it arguably the best alternative for startups and businesses looking to scale while keeping costs low.

The reason why DeepSeek can keep up the affordability is explained well in this Reddit post. Unlike the other closed-source organizations mentioned above, DeepSeek offers an open-source approach, allowing developers to fine-tune and deploy the models according to their specific needs, only paying for hardware usage at the end of the day.

Moreover, the benchmarking results are simply astonishing!

Why NOT DeepSeek

Despite visibly being better at handling reasoning and being so fast, DeepSeek is giving the world a window into Chinese censorship and information control. Both the models conveniently appear to censor answers on sensitive Chinese topics, a rather common practice seen in mainland China. This article gives an overview in this regard.

Is DeepSeek Production Ready?

Deploying DeepSeek in production environments would require better, uncensored reliability guarantees and security measures against biases, areas where DeepSeek is heavily lagging. This raises concerns about governance, safety, and responsible AI deployment.

Inability to Support Function Calling

Another one of DeepSeek’s most critical shortcomings is its lack of built-in function calling capabilities. Modern AI apps, such as automation and agent-based workflows, heavily rely on function calling for API interactions and structured outputs. Without this feature, developers may find it challenging to integrate DeepSeek into systems requiring dynamic, multi-step processing.

Ending Notes

DeepSeek sets new standards for the AI industry, but its production readiness is still up for debate. While it offers an affordable, open-source alternative to popular LLMs enthusiasts use and love, its current limitations and of course censorship make it a rather risky choice for end-user applications. Developers must weigh the pointers above carefully before committing to DeepSeek for production workloads. For now, DeepSeek looks like a WIP.