I've been following the AI market since the release of ChatGPT and use it in my day-to-day work (IT, development).
Of course, I tested and tried to use DeepSeek as well, but as I wrote earlier - I didn't notice any advantages, but found clear disadvantages. I am not ready to confirm the words that DeepSeek used a set of pre-trained training data from ChatGPT “stock”, but the gap in the relevance of the data is very noticeable. From the point of view of reasoning - DeepSeek did not impress me either, because there is experience of practical work with claude, for example, where the quality of reasoning, work with the support of dialog chronology and data, is noticeably higher.
Of course, I tested and tried to use DeepSeek as well, but as I wrote earlier - I didn't notice any advantages, but found clear disadvantages. I am not ready to confirm the words that DeepSeek used a set of pre-trained training data from ChatGPT “stock”, but the gap in the relevance of the data is very noticeable. From the point of view of reasoning - DeepSeek did not impress me either, because there is experience of practical work with claude, for example, where the quality of reasoning, work with the support of dialog chronology and data, is noticeably higher.
To compare performance, I tested DeepSeek and ChatGPT to create signature in bbcode. ChatGPT was not able to create according to my instructions but DeepSeek crashed in the process which is quite surprising. Since you have more experience working with AI tools, you are in better position to judgement which one is better at the moment.
From my perspective, DeepSeek got the hype because of there claim that it's built in 6 million $ budget but if it's performance doesn't match that of ChatGPT then game might be over for DeepSeek very soon.
Between thanks for introducing me to Claude.
If you compare DeepSeek and ChatGPT, I will answer honestly - of these two options is better..... Claude
For IT tasks, it is really superior to both of the options you suggested. Plus - very high quality system of reasoning, conclusions, logical thinking.
Regarding the “6 million dollar budget” there have already been a lot of explanations that reveal, let's say, not quite a fair assessment of this figure, plus the origin of the “prepared data” that is the basis for such systems. In short, and to put it mildly, it is not their data
So both in terms of professional benefits and moral and ethical nuances of DeepSeek's origin, I will definitely not use it and support it with my money