OpenAI launches o3-Mini as it fights back against DeepSeek

Artificial intelligence firm OpenAI has taken the initiative to fight back against the DeepSeek R1 model, announcing the launch of its o3-Mini. The company moved swiftly on Friday in an attempt to respond directly to the Chinese startup’s AI model that has got the market talk.

The DeepSeek R1 model has been hailed as a next level AI model, with reports claiming that the model produces fantastic responses, mimicking the human thought process. The model also reduces memory usage and is more cost-effective than other competitors in the market. Aside from its incredible performance, the making of the model reportedly cost $6 million, compared to OpenAI’s ChatGPT-4 which costs close to $100 million.

OpenAI calls o3-Mini the ‘most cost-efficient’ model

With DeepSeek’s recent model being hailed as the most cost-efficient in the market, it may have struck a chord, pushing OpenAI to act. According to its blog post, the company mentioned that the o3-Mini model is the ‘most cost-efficient’ in their reasoning series. “We’re releasing OpenAI o3-mini, the newest, most cost-efficient model in our reasoning series, available in both ChatGPT and the API today,” the blog post read.

In its statement, the company also teased some of the capabilities of the new model, noting that it advances what smaller models can achieve. “Previewed in December 2024⁠, this powerful and fast model advances the boundaries of what small models can achieve (…) all while maintaining the low cost and reduced latency of OpenAI o1-mini,” OpenAI added.

According to its statement, the artificial intelligence company will allow users to enjoy the reasoning capabilities of the models for free for the first time, while allowing paid users to enjoy three times the daily message limit, pushing it from 50 to 150. The firm is opening this up to boost the usage of the new line of reasoning models. The new model, which comes in low, medium, and high versions, will provide users with better answers for more inference.

The company’s “o” family of AI models focuses majorly on reasoning tasks. This means that they are less creative compared to the GPT family models, but they have a chain of thought reasoning included in their models. This allows them to solve complex problems, build better structure codes, and backtrack on wrong analysis. While the new model is worse than the o1-Mini in general knowledge and multilingual chain of thought, it performs better at coding, factuality, and every other benchmark.

DeepSeek breakthrough and its effects

The breakthrough achieved by DeepSeek with its AI model, which delivered better results than OpenAI’s AI model, caused a bit of a stir in the United States tech industry. Aside from the buzz, the breakthrough also wiped off about $1 trillion from the US tech market, with Nvidia shedding $600 billion in market value. The development has also called into question the demand for its expensive AI chips in the future.

The efficiency gap is coming from the novel approach that DeeepSeek took in its model architecture. While artificial companies in the US were trying to increase the computing power of their models, DeepSeek was able to streamline how the models process information, leading to more efficiency. The pressure was cranked up a notch when Chinese tech company Alibaba launched its Qwen2.5 Max.

Meanwhile, o3-Mini will be looking at increasing the gap, with the model running 24% faster than the previous model. It also matches some older models and beats others on key benchmarks. While its pricing is quite competitive, it is higher than that of DeepSeek’s R1. The o3-Mini is priced at $0.55 per million input tokens, and $4.40 per million output tokens, while DeepSeek is priced at $0.14 and $2.19 for the same volumes.

Cryptopolitan Academy: How to Write a Web3 Resume That Lands Interviews - FREE Cheat Sheet