DeepSeek R2 reduced costs by 97 per cent!

China-based artificial intelligence company DeepSeek will make a strong breakthrough in the sector with its new generation model R2. According to leaked information, DeepSeek R2 doubles its previous generation performance, reaching 1.2 trillion operations per second.

DeepSeek R2 reduces costs by 97 per cent

The new model also makes a big difference on the cost side, offering 97 per cent lower transaction costs. DeepSeek R2 is capable of competing directly with top AI models such as GPT-4 Turbo and Google Gemini 2.0 Pro.

In the new version, the cost per million tokens per input is set at $0.07, and the cost per million tokens per production is set at $0.27. This data makes DeepSeek R2 one of the lowest cost AI solutions on the market.

An important change is also on the hardware side. DeepSeek R2 greatly reduces its dependence on Nvidia accelerators; 82 per cent uses Huawei Ascend 910B accelerators.

In this way, the new embargoes imposed by the US are prevented from affecting performance. In addition, with vertical integration, tighter control is provided on both hardware and software sides.

DeepSeek’s new model will be one of China’s steps to reduce foreign dependence on artificial intelligence technologies. With the introduction of DeepSeek R2 to the market, a new balance is expected to emerge in the global artificial intelligence competition.

So what do you think about this issue? You can share your opinions with us in the comments section below.