Chinese AI Startup DeepSeek Unveils New Open-source Reasoning Model DeepSeek-R1

www.analyticsdrift.com

Image source: Analytics Drift

DeepSeek, a leading Chinese AI startup, has released DeepSeek-R1, an open-source reasoning model under the MIT license.

Image source: Deepseek

DeepSeek Introduces DeepSeek-R1

DeepSeek-R1 is designed to solve complex mathematical and analytical problems. It consists of two core versions: DeepSeek-R1-Zero and DeepSeek-R1.

Image source: Deepseek

DeepSeek-R1: Two Core Versions

DeepSeek-R1 has performed well on several benchmarks, including AIME 2024, Codeforce, MMLU, GPQA Diamond, and AlpacaEval 2.0, transcending OpenAI’s o1 model.

Image source: Deepseek

Challenge to the Dominance of OpenAI

The ability to generate stepwise responses makes DeepSeek-R1 suitable for solving complex problems in research, increasing enthusiasm amongst scientists about its usage.

Image source: Canva

Stepwise Solution Fuels Interests of Scientists

While enterprises can use it for free due to its open-source feature, additional expenses may occur for customization and fine-tuning.

Image source: Canva

Free Base Model, Customization Costs May Apply

Businesses are also reluctant to integrate DeepSeek-R1 in their workflows owing to a lack of clarity on AI regulations and geopolitical reasons.

Image source: Canva

Impact of Regulatory and Geopolitical Factors