Can DeepSeek R1 Take On OpenAI o1? Benchmarks Say Yes

Why Trust Techopedia

As time goes on, artificial intelligence is increasingly going open source.

New to the playing field is DeepSeek, which released a powerful large language model (LLM) in January 2025 that displays “remarkable reasoning capabilities” and performance comparable to OpenAI o1 on AI reasoning tasks.

DeepSeek, a Chinese company founded by Liang Wenfang in 2023, demonstrates with its model that open-source AI can compete with proprietary AI products and even surpass them in key areas, especially in the hot topic of price.

That said — a discount price can come with risks, with emerging reports that DeepSeek is leaking user data. ‘Caveat emptor’, as they say.

But can DeepSeek take on OpenAI? If R1 is any indicator, this Hangzhou-based startup will be a key player in the generative AI market for the foreseeable future. Here’s why.

Key Takeaways

  • In January 2025 DeepSeek Launched R1, a powerful LLM with 671 billion parameters,
  • The model uses techniques like reinforcement learning and chain-of-thought reasoning to improve the accuracy of responses.
  • DeepSeek’s model is so powerful that it displays performance comparable to OpenAI o1 on reasoning tasks.
  • Experts have reacted positively to R1’s open-source approach and low overall cost.

Everything We Know About DeepSeek R1 So Far

DeepSeek R1 is an open source AI model, which according to benchmarks performs on par with OpenAI across mathematics, coding, and reasoning tasks. The model is available via the company’s official website or via API.

Advertisements

According to DeepSeek’s research paper, R1 has been trained using techniques like reinforcement learning and chain of thought reasoning to increase the accuracy of the model’s responses.

One of R1’s core selling points, besides its open-source availability, is the low cost of its API. DeepSeek R1 is available via API for $0.14 per million input tokens and $2.19 per million output tokens.

In contrast, o1 is available for $15 per million input tokens and $60 per 1 million output tokens, which makes R1 extraordinarily cheaper than running o1 with comparable performance.

That said, after a few weeks of performance scrutiny, the team at Wiz Research found gaping holes in DeepSeek’s security, with an exposed database revealing chat prompts and internal data.

The exposure let the team download millions of lines of log streams, including “highly sensitive information.”

Hypothetically speaking, if OpenAI left such a gap in security, the company would be exposed to lawsuits.

A plaintiff would have an exceedingly difficult, if not impossible, journey to get legal satisfaction chasing an open-source project based in China.

R1 vs o1: Key Performance Metrics

How does R1 compare to o1? Looking at the performance metrics outlined in DeepSeek’s research paper, we see that R1 performs on par with o1 across various mathematics and coding tasks. Below is a brief comparison of the two model’s performance:

Benchmark R1 O1
AIME 2024 mathematics benchmark 79.8% 79.2%
MATH-500 97.3% 96.4%
Codeforces 96.3% 96.6%
MMLU 90.8% 91.8%
GPQA Diamond 71.5% 75.7%
SWE-bench verified 49.2% 48.9%

What’s notable about these scores is how close R1 and o1 perform against each other across the board. On AIME 2024, MATH-500, codeforces, MMLU, and SWE-bench verified the scores are almost identical.

The only benchmark where we see o1 pull away slightly is GPQA Diamond, a set of multiple-choice questions written by domain experts in biology, physics, and chemistry.

Could R1 Threaten OpenAI?

DeepSeek R1 excels in providing an open-source, high-performance LLM at a competitive price point, but the LLM market is so oversaturated that it will be difficult to displace OpenAI.

While there is certainly a lot of interest surrounding R1, it’s competing against products with clearly defined differentiation. For instance, OpenAI’s emphasis on agentic AI, Grok’s less-restrictive content moderation, and Claude’s safety-focused outputs all offer clear differentiation, which DeepSeek has yet to fully define.

At the same time, OpenAI is difficult to challenge due to its immense commercial partnerships. For instance, in January 2025 OpenAI announced The Stargate Project, which will see funders like SoftBank, OpenAI, Oracle, and MGX invest $500 billion into the startup over the next four years.

However, one area where R1 does pose a threat to OpenAI is pricing. Companies that found O1 prohibitively expensive may be drawn to experiment with R1 as a more cost-effective alternative. After all, many companies struggle to meet the cost of training and running LLMs.

Expert Reactions to DeepSeek R1 So Far

The tech and AI community has responded positively to the release of DeepSeek R1, with many praising the solution for its open-source approach and low cost.

“We are living in a timeline where a non-U.S. company is keeping the original mission of OpenAI alive – truly open, frontier research that empowers all…they are perhaps the first OSS project that shows major, sustained growth of an RL flywheel,” Dr Jim Fan, Nvidia senior research manager and cofounder of GEAR Lab, posted on X.

Arnaud Bertrand, founder of HouseTrip, believes that DeepSeek will be a serious competitor to OpenAI.

“Most people probably don’t realize how bad news [sic] China’s Deepseek is for OpenAI. They’ve come up with a model that matches and even exceeds OpenAI’s latest model o1 on various benchmarks, and they’re charging just 3% of the price,” Bertrand said in a post on X.

While it remains to be seen whether R1 will be a thorn in the side of OpenAI, the true winners are open-source researchers, who now have a new high-performance model to play with alongside other top performers like Llama 3.

The Bottom Line

DeepSeek can potentially lower the entry barrier to LLM and enrich the AI ecosystem. Whether it can stand up to OpenAI remains to be seen, and whether it can do it safely globally, especially operating in a regime like China, is a whole other question.

Either way, the release of R1 highlights that generative AI moves fast, and we can’t afford to write off new startups emerging to shake up the scene.

FAQs

What is DeepSeek R1?

How does DeepSeek R1 compare to OpenAI o1?

Why is DeepSeek R1 considered important?

What is DeepSeek R1 pricing?

What are the key benchmarks where DeepSeek R1 excels?

Will DeepSeek R1 replace OpenAI o1?

Advertisements

Related Reading

Related Terms

Advertisements
Tim Keary
Technology Writer
Tim Keary
Technology Writer

Tim Keary is a technology writer and reporter covering AI, cybersecurity, and enterprise technology. Before joining Techopedia full-time in 2023, his work appeared on VentureBeat, Forbes Advisor, and other notable technology platforms, where he covered the latest trends and innovations in technology. He holds a Master’s degree in History from the University of Kent, where he learned of the value of breaking complex topics down into simple concepts. Outside of writing and conducting interviews, Tim produces music and trains in Mixed Martial Arts (MMA).