Recently DeepSeek AI was launched and rapidly gained market attention. This is because it is open-source, low-cost, and high-performance. It is SOTA reasoning model which is trained using reinforcement learning directly without supervised tuning.
Basically DeepSeek launched various models in their research paper. Here is the link to the research paper.
One user on X(Twitter) is showing the real DeepSeek R1 in action.
ChatGPT is now dethroned.
— Hasan Toor ✪ (@hasantoxr) January 27, 2025
DeepSeek R1 takes the world by storm, and everyone is talking about it.
Here are 21 wild examples so far: pic.twitter.com/Scu2Pl6b74
DeepSeek AI is highly optimized and give more predictions in tasks like math, coding, thinking etc. The model provides 90% accuracy in any terms.
DeepSeek-R1 achieves 79.8% Pass@1 on AIME 2024, surpassing OpenAI-o1–1217 by a slight margin. On MATH-500, it achieves a high score of 97.3%, comparable to that of OpenAI-o1–1217 and far ahead of the other models.
As DeekSeek is open source, they release several models –
- DeepSeek R1 Zero
- DeepSeek R1
- Six Dense Models
Glad that you read the post, For more such interesting topics, subscribe to our newsletter here.
Also Read, What Is DeepSeek R1 And Why Everyone Is Talking About It: Everything you need to know
