Deepseek’s AI is faster, smarter and cheaper than OpenAI’s models.
- Chinese AI Startup DeepSeek released an upgraded AI Model called V3-0324 on Hugging Face.
- The V3-0324 model offers improved reasoning and programming abilities over its predecessors.
- DeepSeek says its AI models are comparable to or better than those of American AI developers such as OpenAI and Anthropic.
DeepSeek announced a major update to its AI model last week, which is making people as excited as they were when the Hugging Face has now launched the DeepSeek-V3-0324 AI model, which will set up a more intense competition with OpenAI and other AI providers.
According the company’s testing, DeepSeek’s new iteration V3 model boasts measurable improvements in reasoning and coding abilities. This release is notable because it not only improves thinking and coding, but also the speed of improvement.
Since its formation last year, DeepSeek is moving quickly. The original V3 model was released in December. A month later the R1 model was released for more comprehensive research. V3-0324, the model’s release date is March 2024, is now available.
DeepSeek demand
These improvements bring the model up to parity with OpenAI’s GPT-4 and Anthropic’s Claude 2 models. DeepSeek claims that even though they don’t have the same power as OpenAI’s GPT-4 or Anthropic’s Claude 2 models, they are still cheaper.
This is a major selling point, as AI usage and costs continue to rise. OpenAI and Google are able to afford huge cloud budgets, which most companies could not reach without partnerships such as OpenAI’s one with Microsoft. This exclusivity will disappear if DeepSeek achieves more at a lower cost.
The U.S. dominance in AI models is already starting to decline, thanks to Chinese startups such as DeepSeek. It is no longer surprising when the latest AI model comes from Shenzhen and Hangzhou. Geopolitical concerns and business concerns have prompted calls to ban DeepSeek at least from the U.S. Government.
Although you won’t be able to change your schedule tomorrow because of DeepSeek’s newest release, it does hint that the demand for computational power and electricity needed to fuel next-generation AI may not be as high as expected.
This could also mean that the AI chatbot who is rewriting your resume or debugging you website speaks fluent Mandarin.
- I found a surprising difference in the search capabilities of DeepSeek and ChatGPT
- OpenAI asks US government to ban DeepSeek as it is’state subsidized’ and “state controlled”
- Gemini 2.0 Flash vs DeepSeek R1: you might be surprised at the winner
