I know this rattles a lot of people's jimmies but in my own personal benchmarks (three problems a 10 year old could easily solve), all "frontier" LLMs are still getting the same output quality as GPT4 or even GPT3. So at least for me cheapest provider wins.
No, I'm not sharing the problems with you. Unlike a certain influencer who really likes pelicans, I don't get paid for boosting these companies.
In my daily usage and coding ChatGPT 5.2 is far better than 3.5 or 4… but as you prefer.
Anyway the point here is Chinese LLM VS Americans. DeepSeek is very close to Claude or ChatGPT performance, that’s why companies are using it, probably they don’t care about privacy and security issues. That are why I’m not using Chinese LLMs
I know this rattles a lot of people's jimmies but in my own personal benchmarks (three problems a 10 year old could easily solve), all "frontier" LLMs are still getting the same output quality as GPT4 or even GPT3. So at least for me cheapest provider wins. No, I'm not sharing the problems with you. Unlike a certain influencer who really likes pelicans, I don't get paid for boosting these companies.
In my daily usage and coding ChatGPT 5.2 is far better than 3.5 or 4… but as you prefer.
Anyway the point here is Chinese LLM VS Americans. DeepSeek is very close to Claude or ChatGPT performance, that’s why companies are using it, probably they don’t care about privacy and security issues. That are why I’m not using Chinese LLMs