gemini models solve a problem in 80% less tokens so that's something to think about.
Source?
https://help.kagi.com/kagi/ai/llm-benchmark.html
Source?
https://help.kagi.com/kagi/ai/llm-benchmark.html