DeepSeek praised for affordability but lags behind ChatGPT and Claude in effectiveness

Since its launch in late January, DeepSeek has attracted attention for its competitive performance at a lower cost. However, a Business Insider survey of frequent AI users suggests the Chinese-developed model is not yet a reliable alternative for professional tasks.

Sainag Nethala, a 33-year-old technical account manager, tested DeepSeek’s R1 model after its Jan. 20 release but found it less useful for general inquiries. ChatGPT remains his go-to for "random questions and brainstorming," while Claude is best for "more thoughtful analysis," he said.

While users appreciate DeepSeek’s open-source model and low cost, its writing capabilities remain a concern.

Robert Benson-May, an accountant and newsletter writer, relies on paid versions of ChatGPT and Claude for drafting emails, summarizing documents, and creating content. He noted that DeepSeek offers more creativity and fewer content restrictions than its competitors but still prefers Claude for writing, as it aligns better with his personal style.

DeepSeek app on a smartphone. Photo from Pexels

For programmers, DeepSeek’s biggest advantage is its open-source nature, allowing customization. Ankit Anchlia, a 37-year-old software engineer in Austin, U.S., said he "fine-tunes" AI models for specific needs but still uses free versions of ChatGPT and Claude for coding, which can save him minutes to an hour per task.

Despite its strengths, some professionals remain unimpressed.

New Jersey realtor James Crisp, 41, who uses ChatGPT for marketing copy and real estate listings, found R1 struggled with industry-specific terminology. "It had a hard time building out a listing description for one of my properties because it doesn't use the proper Western terminology," he said.

While DeepSeek’s free access is appealing, he does not plan to switch. He said DeepSeek is a good tool for those who want a free AI and do not mind "proofreading everything," but for professionals with "tailored AI models," it offers "no true value."

Tina Willis, an Orlando-based car accident and injury lawyer, tested R1 against ChatGPT and Claude for drafting legal documents. She found DeepSeek’s output less detailed and accurate.

"I probably won't be back to DeepSeek anytime soon," she said.